Linux Audio

Check our new training course

Loading...
v5.14.15
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * Copyright (c) 2007-2017 Nicira, Inc.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   4 */
   5
   6#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   7
   8#include "flow.h"
   9#include "datapath.h"
  10#include <linux/uaccess.h>
  11#include <linux/netdevice.h>
  12#include <linux/etherdevice.h>
  13#include <linux/if_ether.h>
  14#include <linux/if_vlan.h>
  15#include <net/llc_pdu.h>
  16#include <linux/kernel.h>
  17#include <linux/jhash.h>
  18#include <linux/jiffies.h>
  19#include <linux/llc.h>
  20#include <linux/module.h>
  21#include <linux/in.h>
  22#include <linux/rcupdate.h>
  23#include <linux/if_arp.h>
  24#include <linux/ip.h>
  25#include <linux/ipv6.h>
  26#include <linux/sctp.h>
  27#include <linux/tcp.h>
  28#include <linux/udp.h>
  29#include <linux/icmp.h>
  30#include <linux/icmpv6.h>
  31#include <linux/rculist.h>
  32#include <net/geneve.h>
  33#include <net/ip.h>
  34#include <net/ipv6.h>
  35#include <net/ndisc.h>
  36#include <net/mpls.h>
  37#include <net/vxlan.h>
  38#include <net/tun_proto.h>
  39#include <net/erspan.h>
  40
  41#include "flow_netlink.h"
  42
  43struct ovs_len_tbl {
  44	int len;
  45	const struct ovs_len_tbl *next;
  46};
  47
  48#define OVS_ATTR_NESTED -1
  49#define OVS_ATTR_VARIABLE -2
  50
  51static bool actions_may_change_flow(const struct nlattr *actions)
  52{
  53	struct nlattr *nla;
  54	int rem;
  55
  56	nla_for_each_nested(nla, actions, rem) {
  57		u16 action = nla_type(nla);
  58
  59		switch (action) {
  60		case OVS_ACTION_ATTR_OUTPUT:
  61		case OVS_ACTION_ATTR_RECIRC:
  62		case OVS_ACTION_ATTR_TRUNC:
  63		case OVS_ACTION_ATTR_USERSPACE:
  64			break;
  65
  66		case OVS_ACTION_ATTR_CT:
  67		case OVS_ACTION_ATTR_CT_CLEAR:
  68		case OVS_ACTION_ATTR_HASH:
  69		case OVS_ACTION_ATTR_POP_ETH:
  70		case OVS_ACTION_ATTR_POP_MPLS:
  71		case OVS_ACTION_ATTR_POP_NSH:
  72		case OVS_ACTION_ATTR_POP_VLAN:
  73		case OVS_ACTION_ATTR_PUSH_ETH:
  74		case OVS_ACTION_ATTR_PUSH_MPLS:
  75		case OVS_ACTION_ATTR_PUSH_NSH:
  76		case OVS_ACTION_ATTR_PUSH_VLAN:
  77		case OVS_ACTION_ATTR_SAMPLE:
  78		case OVS_ACTION_ATTR_SET:
  79		case OVS_ACTION_ATTR_SET_MASKED:
  80		case OVS_ACTION_ATTR_METER:
  81		case OVS_ACTION_ATTR_CHECK_PKT_LEN:
  82		case OVS_ACTION_ATTR_ADD_MPLS:
  83		case OVS_ACTION_ATTR_DEC_TTL:
  84		default:
  85			return true;
  86		}
  87	}
  88	return false;
  89}
  90
  91static void update_range(struct sw_flow_match *match,
  92			 size_t offset, size_t size, bool is_mask)
  93{
  94	struct sw_flow_key_range *range;
  95	size_t start = rounddown(offset, sizeof(long));
  96	size_t end = roundup(offset + size, sizeof(long));
  97
  98	if (!is_mask)
  99		range = &match->range;
 100	else
 101		range = &match->mask->range;
 102
 103	if (range->start == range->end) {
 104		range->start = start;
 105		range->end = end;
 106		return;
 107	}
 108
 109	if (range->start > start)
 110		range->start = start;
 111
 112	if (range->end < end)
 113		range->end = end;
 114}
 115
 116#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
 117	do { \
 118		update_range(match, offsetof(struct sw_flow_key, field),    \
 119			     sizeof((match)->key->field), is_mask);	    \
 120		if (is_mask)						    \
 121			(match)->mask->key.field = value;		    \
 122		else							    \
 123			(match)->key->field = value;		            \
 124	} while (0)
 125
 126#define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask)	    \
 127	do {								    \
 128		update_range(match, offset, len, is_mask);		    \
 129		if (is_mask)						    \
 130			memcpy((u8 *)&(match)->mask->key + offset, value_p, \
 131			       len);					   \
 132		else							    \
 133			memcpy((u8 *)(match)->key + offset, value_p, len);  \
 134	} while (0)
 135
 136#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask)		      \
 137	SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
 138				  value_p, len, is_mask)
 139
 140#define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask)		    \
 141	do {								    \
 142		update_range(match, offsetof(struct sw_flow_key, field),    \
 143			     sizeof((match)->key->field), is_mask);	    \
 144		if (is_mask)						    \
 145			memset((u8 *)&(match)->mask->key.field, value,      \
 146			       sizeof((match)->mask->key.field));	    \
 147		else							    \
 148			memset((u8 *)&(match)->key->field, value,           \
 149			       sizeof((match)->key->field));                \
 150	} while (0)
 151
 152static bool match_validate(const struct sw_flow_match *match,
 153			   u64 key_attrs, u64 mask_attrs, bool log)
 154{
 155	u64 key_expected = 0;
 156	u64 mask_allowed = key_attrs;  /* At most allow all key attributes */
 157
 158	/* The following mask attributes allowed only if they
 159	 * pass the validation tests. */
 160	mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
 161			| (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
 162			| (1 << OVS_KEY_ATTR_IPV6)
 163			| (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
 164			| (1 << OVS_KEY_ATTR_TCP)
 165			| (1 << OVS_KEY_ATTR_TCP_FLAGS)
 166			| (1 << OVS_KEY_ATTR_UDP)
 167			| (1 << OVS_KEY_ATTR_SCTP)
 168			| (1 << OVS_KEY_ATTR_ICMP)
 169			| (1 << OVS_KEY_ATTR_ICMPV6)
 170			| (1 << OVS_KEY_ATTR_ARP)
 171			| (1 << OVS_KEY_ATTR_ND)
 172			| (1 << OVS_KEY_ATTR_MPLS)
 173			| (1 << OVS_KEY_ATTR_NSH));
 174
 175	/* Always allowed mask fields. */
 176	mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
 177		       | (1 << OVS_KEY_ATTR_IN_PORT)
 178		       | (1 << OVS_KEY_ATTR_ETHERTYPE));
 179
 180	/* Check key attributes. */
 181	if (match->key->eth.type == htons(ETH_P_ARP)
 182			|| match->key->eth.type == htons(ETH_P_RARP)) {
 183		key_expected |= 1 << OVS_KEY_ATTR_ARP;
 184		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
 185			mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
 186	}
 187
 188	if (eth_p_mpls(match->key->eth.type)) {
 189		key_expected |= 1 << OVS_KEY_ATTR_MPLS;
 190		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
 191			mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
 192	}
 193
 194	if (match->key->eth.type == htons(ETH_P_IP)) {
 195		key_expected |= 1 << OVS_KEY_ATTR_IPV4;
 196		if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
 197			mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
 198			mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
 199		}
 200
 201		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
 202			if (match->key->ip.proto == IPPROTO_UDP) {
 203				key_expected |= 1 << OVS_KEY_ATTR_UDP;
 204				if (match->mask && (match->mask->key.ip.proto == 0xff))
 205					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
 206			}
 207
 208			if (match->key->ip.proto == IPPROTO_SCTP) {
 209				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
 210				if (match->mask && (match->mask->key.ip.proto == 0xff))
 211					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
 212			}
 213
 214			if (match->key->ip.proto == IPPROTO_TCP) {
 215				key_expected |= 1 << OVS_KEY_ATTR_TCP;
 216				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 217				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
 218					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
 219					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 220				}
 221			}
 222
 223			if (match->key->ip.proto == IPPROTO_ICMP) {
 224				key_expected |= 1 << OVS_KEY_ATTR_ICMP;
 225				if (match->mask && (match->mask->key.ip.proto == 0xff))
 226					mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
 227			}
 228		}
 229	}
 230
 231	if (match->key->eth.type == htons(ETH_P_IPV6)) {
 232		key_expected |= 1 << OVS_KEY_ATTR_IPV6;
 233		if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
 234			mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
 235			mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
 236		}
 237
 238		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
 239			if (match->key->ip.proto == IPPROTO_UDP) {
 240				key_expected |= 1 << OVS_KEY_ATTR_UDP;
 241				if (match->mask && (match->mask->key.ip.proto == 0xff))
 242					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
 243			}
 244
 245			if (match->key->ip.proto == IPPROTO_SCTP) {
 246				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
 247				if (match->mask && (match->mask->key.ip.proto == 0xff))
 248					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
 249			}
 250
 251			if (match->key->ip.proto == IPPROTO_TCP) {
 252				key_expected |= 1 << OVS_KEY_ATTR_TCP;
 253				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 254				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
 255					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
 256					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 257				}
 258			}
 259
 260			if (match->key->ip.proto == IPPROTO_ICMPV6) {
 261				key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
 262				if (match->mask && (match->mask->key.ip.proto == 0xff))
 263					mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
 264
 265				if (match->key->tp.src ==
 266						htons(NDISC_NEIGHBOUR_SOLICITATION) ||
 267				    match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
 268					key_expected |= 1 << OVS_KEY_ATTR_ND;
 269					/* Original direction conntrack tuple
 270					 * uses the same space as the ND fields
 271					 * in the key, so both are not allowed
 272					 * at the same time.
 273					 */
 274					mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
 275					if (match->mask && (match->mask->key.tp.src == htons(0xff)))
 276						mask_allowed |= 1 << OVS_KEY_ATTR_ND;
 277				}
 278			}
 279		}
 280	}
 281
 282	if (match->key->eth.type == htons(ETH_P_NSH)) {
 283		key_expected |= 1 << OVS_KEY_ATTR_NSH;
 284		if (match->mask &&
 285		    match->mask->key.eth.type == htons(0xffff)) {
 286			mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
 287		}
 288	}
 289
 290	if ((key_attrs & key_expected) != key_expected) {
 291		/* Key attributes check failed. */
 292		OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
 293			  (unsigned long long)key_attrs,
 294			  (unsigned long long)key_expected);
 295		return false;
 296	}
 297
 298	if ((mask_attrs & mask_allowed) != mask_attrs) {
 299		/* Mask attributes check failed. */
 300		OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
 301			  (unsigned long long)mask_attrs,
 302			  (unsigned long long)mask_allowed);
 303		return false;
 304	}
 305
 306	return true;
 307}
 308
 309size_t ovs_tun_key_attr_size(void)
 310{
 311	/* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
 312	 * updating this function.
 313	 */
 314	return    nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
 315		+ nla_total_size(16)   /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
 316		+ nla_total_size(16)   /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
 317		+ nla_total_size(1)    /* OVS_TUNNEL_KEY_ATTR_TOS */
 318		+ nla_total_size(1)    /* OVS_TUNNEL_KEY_ATTR_TTL */
 319		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
 320		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_CSUM */
 321		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_OAM */
 322		+ nla_total_size(256)  /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
 323		/* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
 324		 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
 325		 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
 326		 */
 327		+ nla_total_size(2)    /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
 328		+ nla_total_size(2);   /* OVS_TUNNEL_KEY_ATTR_TP_DST */
 329}
 330
 331static size_t ovs_nsh_key_attr_size(void)
 332{
 333	/* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
 334	 * updating this function.
 335	 */
 336	return  nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
 337		/* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
 338		 * mutually exclusive, so the bigger one can cover
 339		 * the small one.
 340		 */
 341		+ nla_total_size(NSH_CTX_HDRS_MAX_LEN);
 342}
 343
 344size_t ovs_key_attr_size(void)
 345{
 346	/* Whenever adding new OVS_KEY_ FIELDS, we should consider
 347	 * updating this function.
 348	 */
 349	BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 29);
 350
 351	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
 352		+ nla_total_size(0)   /* OVS_KEY_ATTR_TUNNEL */
 353		  + ovs_tun_key_attr_size()
 354		+ nla_total_size(4)   /* OVS_KEY_ATTR_IN_PORT */
 355		+ nla_total_size(4)   /* OVS_KEY_ATTR_SKB_MARK */
 356		+ nla_total_size(4)   /* OVS_KEY_ATTR_DP_HASH */
 357		+ nla_total_size(4)   /* OVS_KEY_ATTR_RECIRC_ID */
 358		+ nla_total_size(4)   /* OVS_KEY_ATTR_CT_STATE */
 359		+ nla_total_size(2)   /* OVS_KEY_ATTR_CT_ZONE */
 360		+ nla_total_size(4)   /* OVS_KEY_ATTR_CT_MARK */
 361		+ nla_total_size(16)  /* OVS_KEY_ATTR_CT_LABELS */
 362		+ nla_total_size(40)  /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
 363		+ nla_total_size(0)   /* OVS_KEY_ATTR_NSH */
 364		  + ovs_nsh_key_attr_size()
 365		+ nla_total_size(12)  /* OVS_KEY_ATTR_ETHERNET */
 366		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
 367		+ nla_total_size(4)   /* OVS_KEY_ATTR_VLAN */
 368		+ nla_total_size(0)   /* OVS_KEY_ATTR_ENCAP */
 369		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
 370		+ nla_total_size(40)  /* OVS_KEY_ATTR_IPV6 */
 371		+ nla_total_size(2)   /* OVS_KEY_ATTR_ICMPV6 */
 372		+ nla_total_size(28); /* OVS_KEY_ATTR_ND */
 373}
 374
 375static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
 376	[OVS_VXLAN_EXT_GBP]	    = { .len = sizeof(u32) },
 377};
 378
 379static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
 380	[OVS_TUNNEL_KEY_ATTR_ID]	    = { .len = sizeof(u64) },
 381	[OVS_TUNNEL_KEY_ATTR_IPV4_SRC]	    = { .len = sizeof(u32) },
 382	[OVS_TUNNEL_KEY_ATTR_IPV4_DST]	    = { .len = sizeof(u32) },
 383	[OVS_TUNNEL_KEY_ATTR_TOS]	    = { .len = 1 },
 384	[OVS_TUNNEL_KEY_ATTR_TTL]	    = { .len = 1 },
 385	[OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
 386	[OVS_TUNNEL_KEY_ATTR_CSUM]	    = { .len = 0 },
 387	[OVS_TUNNEL_KEY_ATTR_TP_SRC]	    = { .len = sizeof(u16) },
 388	[OVS_TUNNEL_KEY_ATTR_TP_DST]	    = { .len = sizeof(u16) },
 389	[OVS_TUNNEL_KEY_ATTR_OAM]	    = { .len = 0 },
 390	[OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS]   = { .len = OVS_ATTR_VARIABLE },
 391	[OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS]    = { .len = OVS_ATTR_NESTED,
 392						.next = ovs_vxlan_ext_key_lens },
 393	[OVS_TUNNEL_KEY_ATTR_IPV6_SRC]      = { .len = sizeof(struct in6_addr) },
 394	[OVS_TUNNEL_KEY_ATTR_IPV6_DST]      = { .len = sizeof(struct in6_addr) },
 395	[OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS]   = { .len = OVS_ATTR_VARIABLE },
 396	[OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE]   = { .len = 0 },
 397};
 398
 399static const struct ovs_len_tbl
 400ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
 401	[OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
 402	[OVS_NSH_KEY_ATTR_MD1]  = { .len = sizeof(struct ovs_nsh_key_md1) },
 403	[OVS_NSH_KEY_ATTR_MD2]  = { .len = OVS_ATTR_VARIABLE },
 404};
 405
 406/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
 407static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
 408	[OVS_KEY_ATTR_ENCAP]	 = { .len = OVS_ATTR_NESTED },
 409	[OVS_KEY_ATTR_PRIORITY]	 = { .len = sizeof(u32) },
 410	[OVS_KEY_ATTR_IN_PORT]	 = { .len = sizeof(u32) },
 411	[OVS_KEY_ATTR_SKB_MARK]	 = { .len = sizeof(u32) },
 412	[OVS_KEY_ATTR_ETHERNET]	 = { .len = sizeof(struct ovs_key_ethernet) },
 413	[OVS_KEY_ATTR_VLAN]	 = { .len = sizeof(__be16) },
 414	[OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
 415	[OVS_KEY_ATTR_IPV4]	 = { .len = sizeof(struct ovs_key_ipv4) },
 416	[OVS_KEY_ATTR_IPV6]	 = { .len = sizeof(struct ovs_key_ipv6) },
 417	[OVS_KEY_ATTR_TCP]	 = { .len = sizeof(struct ovs_key_tcp) },
 418	[OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
 419	[OVS_KEY_ATTR_UDP]	 = { .len = sizeof(struct ovs_key_udp) },
 420	[OVS_KEY_ATTR_SCTP]	 = { .len = sizeof(struct ovs_key_sctp) },
 421	[OVS_KEY_ATTR_ICMP]	 = { .len = sizeof(struct ovs_key_icmp) },
 422	[OVS_KEY_ATTR_ICMPV6]	 = { .len = sizeof(struct ovs_key_icmpv6) },
 423	[OVS_KEY_ATTR_ARP]	 = { .len = sizeof(struct ovs_key_arp) },
 424	[OVS_KEY_ATTR_ND]	 = { .len = sizeof(struct ovs_key_nd) },
 425	[OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
 426	[OVS_KEY_ATTR_DP_HASH]	 = { .len = sizeof(u32) },
 427	[OVS_KEY_ATTR_TUNNEL]	 = { .len = OVS_ATTR_NESTED,
 428				     .next = ovs_tunnel_key_lens, },
 429	[OVS_KEY_ATTR_MPLS]	 = { .len = OVS_ATTR_VARIABLE },
 430	[OVS_KEY_ATTR_CT_STATE]	 = { .len = sizeof(u32) },
 431	[OVS_KEY_ATTR_CT_ZONE]	 = { .len = sizeof(u16) },
 432	[OVS_KEY_ATTR_CT_MARK]	 = { .len = sizeof(u32) },
 433	[OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
 434	[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
 435		.len = sizeof(struct ovs_key_ct_tuple_ipv4) },
 436	[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
 437		.len = sizeof(struct ovs_key_ct_tuple_ipv6) },
 438	[OVS_KEY_ATTR_NSH]       = { .len = OVS_ATTR_NESTED,
 439				     .next = ovs_nsh_key_attr_lens, },
 440};
 441
 442static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
 443{
 444	return expected_len == attr_len ||
 445	       expected_len == OVS_ATTR_NESTED ||
 446	       expected_len == OVS_ATTR_VARIABLE;
 447}
 448
 449static bool is_all_zero(const u8 *fp, size_t size)
 450{
 451	int i;
 452
 453	if (!fp)
 454		return false;
 455
 456	for (i = 0; i < size; i++)
 457		if (fp[i])
 458			return false;
 459
 460	return true;
 461}
 462
 463static int __parse_flow_nlattrs(const struct nlattr *attr,
 464				const struct nlattr *a[],
 465				u64 *attrsp, bool log, bool nz)
 466{
 467	const struct nlattr *nla;
 468	u64 attrs;
 469	int rem;
 470
 471	attrs = *attrsp;
 472	nla_for_each_nested(nla, attr, rem) {
 473		u16 type = nla_type(nla);
 474		int expected_len;
 475
 476		if (type > OVS_KEY_ATTR_MAX) {
 477			OVS_NLERR(log, "Key type %d is out of range max %d",
 478				  type, OVS_KEY_ATTR_MAX);
 479			return -EINVAL;
 480		}
 481
 482		if (attrs & (1 << type)) {
 483			OVS_NLERR(log, "Duplicate key (type %d).", type);
 484			return -EINVAL;
 485		}
 486
 487		expected_len = ovs_key_lens[type].len;
 488		if (!check_attr_len(nla_len(nla), expected_len)) {
 489			OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
 490				  type, nla_len(nla), expected_len);
 491			return -EINVAL;
 492		}
 493
 494		if (!nz || !is_all_zero(nla_data(nla), nla_len(nla))) {
 495			attrs |= 1 << type;
 496			a[type] = nla;
 497		}
 498	}
 499	if (rem) {
 500		OVS_NLERR(log, "Message has %d unknown bytes.", rem);
 501		return -EINVAL;
 502	}
 503
 504	*attrsp = attrs;
 505	return 0;
 506}
 507
 508static int parse_flow_mask_nlattrs(const struct nlattr *attr,
 509				   const struct nlattr *a[], u64 *attrsp,
 510				   bool log)
 511{
 512	return __parse_flow_nlattrs(attr, a, attrsp, log, true);
 513}
 514
 515int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
 516		       u64 *attrsp, bool log)
 517{
 518	return __parse_flow_nlattrs(attr, a, attrsp, log, false);
 519}
 520
 521static int genev_tun_opt_from_nlattr(const struct nlattr *a,
 522				     struct sw_flow_match *match, bool is_mask,
 523				     bool log)
 524{
 525	unsigned long opt_key_offset;
 526
 527	if (nla_len(a) > sizeof(match->key->tun_opts)) {
 528		OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
 529			  nla_len(a), sizeof(match->key->tun_opts));
 530		return -EINVAL;
 531	}
 532
 533	if (nla_len(a) % 4 != 0) {
 534		OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
 535			  nla_len(a));
 536		return -EINVAL;
 537	}
 538
 539	/* We need to record the length of the options passed
 540	 * down, otherwise packets with the same format but
 541	 * additional options will be silently matched.
 542	 */
 543	if (!is_mask) {
 544		SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
 545				false);
 546	} else {
 547		/* This is somewhat unusual because it looks at
 548		 * both the key and mask while parsing the
 549		 * attributes (and by extension assumes the key
 550		 * is parsed first). Normally, we would verify
 551		 * that each is the correct length and that the
 552		 * attributes line up in the validate function.
 553		 * However, that is difficult because this is
 554		 * variable length and we won't have the
 555		 * information later.
 556		 */
 557		if (match->key->tun_opts_len != nla_len(a)) {
 558			OVS_NLERR(log, "Geneve option len %d != mask len %d",
 559				  match->key->tun_opts_len, nla_len(a));
 560			return -EINVAL;
 561		}
 562
 563		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
 564	}
 565
 566	opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
 567	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
 568				  nla_len(a), is_mask);
 569	return 0;
 570}
 571
 572static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
 573				     struct sw_flow_match *match, bool is_mask,
 574				     bool log)
 575{
 576	struct nlattr *a;
 577	int rem;
 578	unsigned long opt_key_offset;
 579	struct vxlan_metadata opts;
 580
 581	BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
 582
 583	memset(&opts, 0, sizeof(opts));
 584	nla_for_each_nested(a, attr, rem) {
 585		int type = nla_type(a);
 586
 587		if (type > OVS_VXLAN_EXT_MAX) {
 588			OVS_NLERR(log, "VXLAN extension %d out of range max %d",
 589				  type, OVS_VXLAN_EXT_MAX);
 590			return -EINVAL;
 591		}
 592
 593		if (!check_attr_len(nla_len(a),
 594				    ovs_vxlan_ext_key_lens[type].len)) {
 595			OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
 596				  type, nla_len(a),
 597				  ovs_vxlan_ext_key_lens[type].len);
 598			return -EINVAL;
 599		}
 600
 601		switch (type) {
 602		case OVS_VXLAN_EXT_GBP:
 603			opts.gbp = nla_get_u32(a);
 604			break;
 605		default:
 606			OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
 607				  type);
 608			return -EINVAL;
 609		}
 610	}
 611	if (rem) {
 612		OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
 613			  rem);
 614		return -EINVAL;
 615	}
 616
 617	if (!is_mask)
 618		SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
 619	else
 620		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
 621
 622	opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
 623	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
 624				  is_mask);
 625	return 0;
 626}
 627
 628static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
 629				      struct sw_flow_match *match, bool is_mask,
 630				      bool log)
 631{
 632	unsigned long opt_key_offset;
 633
 634	BUILD_BUG_ON(sizeof(struct erspan_metadata) >
 635		     sizeof(match->key->tun_opts));
 636
 637	if (nla_len(a) > sizeof(match->key->tun_opts)) {
 638		OVS_NLERR(log, "ERSPAN option length err (len %d, max %zu).",
 639			  nla_len(a), sizeof(match->key->tun_opts));
 640		return -EINVAL;
 641	}
 642
 643	if (!is_mask)
 644		SW_FLOW_KEY_PUT(match, tun_opts_len,
 645				sizeof(struct erspan_metadata), false);
 646	else
 647		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
 648
 649	opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
 650	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
 651				  nla_len(a), is_mask);
 652	return 0;
 653}
 654
 655static int ip_tun_from_nlattr(const struct nlattr *attr,
 656			      struct sw_flow_match *match, bool is_mask,
 657			      bool log)
 658{
 659	bool ttl = false, ipv4 = false, ipv6 = false;
 660	bool info_bridge_mode = false;
 661	__be16 tun_flags = 0;
 662	int opts_type = 0;
 663	struct nlattr *a;
 664	int rem;
 665
 666	nla_for_each_nested(a, attr, rem) {
 667		int type = nla_type(a);
 668		int err;
 669
 670		if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
 671			OVS_NLERR(log, "Tunnel attr %d out of range max %d",
 672				  type, OVS_TUNNEL_KEY_ATTR_MAX);
 673			return -EINVAL;
 674		}
 675
 676		if (!check_attr_len(nla_len(a),
 677				    ovs_tunnel_key_lens[type].len)) {
 678			OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
 679				  type, nla_len(a), ovs_tunnel_key_lens[type].len);
 680			return -EINVAL;
 681		}
 682
 683		switch (type) {
 684		case OVS_TUNNEL_KEY_ATTR_ID:
 685			SW_FLOW_KEY_PUT(match, tun_key.tun_id,
 686					nla_get_be64(a), is_mask);
 687			tun_flags |= TUNNEL_KEY;
 688			break;
 689		case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
 690			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
 691					nla_get_in_addr(a), is_mask);
 692			ipv4 = true;
 693			break;
 694		case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
 695			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
 696					nla_get_in_addr(a), is_mask);
 697			ipv4 = true;
 698			break;
 699		case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
 700			SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
 701					nla_get_in6_addr(a), is_mask);
 702			ipv6 = true;
 703			break;
 704		case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
 705			SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
 706					nla_get_in6_addr(a), is_mask);
 707			ipv6 = true;
 708			break;
 709		case OVS_TUNNEL_KEY_ATTR_TOS:
 710			SW_FLOW_KEY_PUT(match, tun_key.tos,
 711					nla_get_u8(a), is_mask);
 712			break;
 713		case OVS_TUNNEL_KEY_ATTR_TTL:
 714			SW_FLOW_KEY_PUT(match, tun_key.ttl,
 715					nla_get_u8(a), is_mask);
 716			ttl = true;
 717			break;
 718		case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
 719			tun_flags |= TUNNEL_DONT_FRAGMENT;
 720			break;
 721		case OVS_TUNNEL_KEY_ATTR_CSUM:
 722			tun_flags |= TUNNEL_CSUM;
 723			break;
 724		case OVS_TUNNEL_KEY_ATTR_TP_SRC:
 725			SW_FLOW_KEY_PUT(match, tun_key.tp_src,
 726					nla_get_be16(a), is_mask);
 727			break;
 728		case OVS_TUNNEL_KEY_ATTR_TP_DST:
 729			SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
 730					nla_get_be16(a), is_mask);
 731			break;
 732		case OVS_TUNNEL_KEY_ATTR_OAM:
 733			tun_flags |= TUNNEL_OAM;
 734			break;
 735		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
 736			if (opts_type) {
 737				OVS_NLERR(log, "Multiple metadata blocks provided");
 738				return -EINVAL;
 739			}
 740
 741			err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
 742			if (err)
 743				return err;
 744
 745			tun_flags |= TUNNEL_GENEVE_OPT;
 746			opts_type = type;
 747			break;
 748		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
 749			if (opts_type) {
 750				OVS_NLERR(log, "Multiple metadata blocks provided");
 751				return -EINVAL;
 752			}
 753
 754			err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
 755			if (err)
 756				return err;
 757
 758			tun_flags |= TUNNEL_VXLAN_OPT;
 759			opts_type = type;
 760			break;
 761		case OVS_TUNNEL_KEY_ATTR_PAD:
 762			break;
 763		case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
 764			if (opts_type) {
 765				OVS_NLERR(log, "Multiple metadata blocks provided");
 766				return -EINVAL;
 767			}
 768
 769			err = erspan_tun_opt_from_nlattr(a, match, is_mask,
 770							 log);
 771			if (err)
 772				return err;
 773
 774			tun_flags |= TUNNEL_ERSPAN_OPT;
 775			opts_type = type;
 776			break;
 777		case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE:
 778			info_bridge_mode = true;
 779			ipv4 = true;
 780			break;
 781		default:
 782			OVS_NLERR(log, "Unknown IP tunnel attribute %d",
 783				  type);
 784			return -EINVAL;
 785		}
 786	}
 787
 788	SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
 789	if (is_mask)
 790		SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
 791	else
 792		SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
 793				false);
 794
 795	if (rem > 0) {
 796		OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
 797			  rem);
 798		return -EINVAL;
 799	}
 800
 801	if (ipv4 && ipv6) {
 802		OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
 803		return -EINVAL;
 804	}
 805
 806	if (!is_mask) {
 807		if (!ipv4 && !ipv6) {
 808			OVS_NLERR(log, "IP tunnel dst address not specified");
 809			return -EINVAL;
 810		}
 811		if (ipv4) {
 812			if (info_bridge_mode) {
 813				if (match->key->tun_key.u.ipv4.src ||
 814				    match->key->tun_key.u.ipv4.dst ||
 815				    match->key->tun_key.tp_src ||
 816				    match->key->tun_key.tp_dst ||
 817				    match->key->tun_key.ttl ||
 818				    match->key->tun_key.tos ||
 819				    tun_flags & ~TUNNEL_KEY) {
 820					OVS_NLERR(log, "IPv4 tun info is not correct");
 821					return -EINVAL;
 822				}
 823			} else if (!match->key->tun_key.u.ipv4.dst) {
 824				OVS_NLERR(log, "IPv4 tunnel dst address is zero");
 825				return -EINVAL;
 826			}
 827		}
 828		if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
 829			OVS_NLERR(log, "IPv6 tunnel dst address is zero");
 830			return -EINVAL;
 831		}
 832
 833		if (!ttl && !info_bridge_mode) {
 834			OVS_NLERR(log, "IP tunnel TTL not specified.");
 835			return -EINVAL;
 836		}
 837	}
 838
 839	return opts_type;
 840}
 841
 842static int vxlan_opt_to_nlattr(struct sk_buff *skb,
 843			       const void *tun_opts, int swkey_tun_opts_len)
 844{
 845	const struct vxlan_metadata *opts = tun_opts;
 846	struct nlattr *nla;
 847
 848	nla = nla_nest_start_noflag(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
 849	if (!nla)
 850		return -EMSGSIZE;
 851
 852	if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
 853		return -EMSGSIZE;
 854
 855	nla_nest_end(skb, nla);
 856	return 0;
 857}
 858
 859static int __ip_tun_to_nlattr(struct sk_buff *skb,
 860			      const struct ip_tunnel_key *output,
 861			      const void *tun_opts, int swkey_tun_opts_len,
 862			      unsigned short tun_proto, u8 mode)
 863{
 864	if (output->tun_flags & TUNNEL_KEY &&
 865	    nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
 866			 OVS_TUNNEL_KEY_ATTR_PAD))
 867		return -EMSGSIZE;
 868
 869	if (mode & IP_TUNNEL_INFO_BRIDGE)
 870		return nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE)
 871		       ? -EMSGSIZE : 0;
 872
 873	switch (tun_proto) {
 874	case AF_INET:
 875		if (output->u.ipv4.src &&
 876		    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
 877				    output->u.ipv4.src))
 878			return -EMSGSIZE;
 879		if (output->u.ipv4.dst &&
 880		    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
 881				    output->u.ipv4.dst))
 882			return -EMSGSIZE;
 883		break;
 884	case AF_INET6:
 885		if (!ipv6_addr_any(&output->u.ipv6.src) &&
 886		    nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
 887				     &output->u.ipv6.src))
 888			return -EMSGSIZE;
 889		if (!ipv6_addr_any(&output->u.ipv6.dst) &&
 890		    nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
 891				     &output->u.ipv6.dst))
 892			return -EMSGSIZE;
 893		break;
 894	}
 895	if (output->tos &&
 896	    nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
 897		return -EMSGSIZE;
 898	if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
 899		return -EMSGSIZE;
 900	if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
 901	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
 902		return -EMSGSIZE;
 903	if ((output->tun_flags & TUNNEL_CSUM) &&
 904	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
 905		return -EMSGSIZE;
 906	if (output->tp_src &&
 907	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
 908		return -EMSGSIZE;
 909	if (output->tp_dst &&
 910	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
 911		return -EMSGSIZE;
 912	if ((output->tun_flags & TUNNEL_OAM) &&
 913	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
 914		return -EMSGSIZE;
 915	if (swkey_tun_opts_len) {
 916		if (output->tun_flags & TUNNEL_GENEVE_OPT &&
 917		    nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
 918			    swkey_tun_opts_len, tun_opts))
 919			return -EMSGSIZE;
 920		else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
 921			 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
 922			return -EMSGSIZE;
 923		else if (output->tun_flags & TUNNEL_ERSPAN_OPT &&
 924			 nla_put(skb, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS,
 925				 swkey_tun_opts_len, tun_opts))
 926			return -EMSGSIZE;
 927	}
 928
 929	return 0;
 930}
 931
 932static int ip_tun_to_nlattr(struct sk_buff *skb,
 933			    const struct ip_tunnel_key *output,
 934			    const void *tun_opts, int swkey_tun_opts_len,
 935			    unsigned short tun_proto, u8 mode)
 936{
 937	struct nlattr *nla;
 938	int err;
 939
 940	nla = nla_nest_start_noflag(skb, OVS_KEY_ATTR_TUNNEL);
 941	if (!nla)
 942		return -EMSGSIZE;
 943
 944	err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
 945				 tun_proto, mode);
 946	if (err)
 947		return err;
 948
 949	nla_nest_end(skb, nla);
 950	return 0;
 951}
 952
 953int ovs_nla_put_tunnel_info(struct sk_buff *skb,
 954			    struct ip_tunnel_info *tun_info)
 955{
 956	return __ip_tun_to_nlattr(skb, &tun_info->key,
 957				  ip_tunnel_info_opts(tun_info),
 958				  tun_info->options_len,
 959				  ip_tunnel_info_af(tun_info), tun_info->mode);
 960}
 961
 962static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
 963				    const struct nlattr *a[],
 964				    bool is_mask, bool inner)
 965{
 966	__be16 tci = 0;
 967	__be16 tpid = 0;
 968
 969	if (a[OVS_KEY_ATTR_VLAN])
 970		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
 971
 972	if (a[OVS_KEY_ATTR_ETHERTYPE])
 973		tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
 974
 975	if (likely(!inner)) {
 976		SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
 977		SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
 978	} else {
 979		SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
 980		SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
 981	}
 982	return 0;
 983}
 984
 985static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
 986				      u64 key_attrs, bool inner,
 987				      const struct nlattr **a, bool log)
 988{
 989	__be16 tci = 0;
 990
 991	if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
 992	      (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
 993	       eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
 994		/* Not a VLAN. */
 995		return 0;
 996	}
 997
 998	if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
 999	      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1000		OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
1001		return -EINVAL;
1002	}
1003
1004	if (a[OVS_KEY_ATTR_VLAN])
1005		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1006
1007	if (!(tci & htons(VLAN_CFI_MASK))) {
1008		if (tci) {
1009			OVS_NLERR(log, "%s TCI does not have VLAN_CFI_MASK bit set.",
1010				  (inner) ? "C-VLAN" : "VLAN");
1011			return -EINVAL;
1012		} else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
1013			/* Corner case for truncated VLAN header. */
1014			OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
1015				  (inner) ? "C-VLAN" : "VLAN");
1016			return -EINVAL;
1017		}
1018	}
1019
1020	return 1;
1021}
1022
1023static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1024					   u64 key_attrs, bool inner,
1025					   const struct nlattr **a, bool log)
1026{
1027	__be16 tci = 0;
1028	__be16 tpid = 0;
1029	bool encap_valid = !!(match->key->eth.vlan.tci &
1030			      htons(VLAN_CFI_MASK));
1031	bool i_encap_valid = !!(match->key->eth.cvlan.tci &
1032				htons(VLAN_CFI_MASK));
1033
1034	if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1035		/* Not a VLAN. */
1036		return 0;
1037	}
1038
1039	if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
1040		OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
1041			  (inner) ? "C-VLAN" : "VLAN");
1042		return -EINVAL;
1043	}
1044
1045	if (a[OVS_KEY_ATTR_VLAN])
1046		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1047
1048	if (a[OVS_KEY_ATTR_ETHERTYPE])
1049		tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1050
1051	if (tpid != htons(0xffff)) {
1052		OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
1053			  (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
1054		return -EINVAL;
1055	}
1056	if (!(tci & htons(VLAN_CFI_MASK))) {
1057		OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_CFI_MASK bit.",
1058			  (inner) ? "C-VLAN" : "VLAN");
1059		return -EINVAL;
1060	}
1061
1062	return 1;
1063}
1064
1065static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1066				     u64 *key_attrs, bool inner,
1067				     const struct nlattr **a, bool is_mask,
1068				     bool log)
1069{
1070	int err;
1071	const struct nlattr *encap;
1072
1073	if (!is_mask)
1074		err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1075						 a, log);
1076	else
1077		err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1078						      a, log);
1079	if (err <= 0)
1080		return err;
1081
1082	err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1083	if (err)
1084		return err;
1085
1086	*key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1087	*key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1088	*key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1089
1090	encap = a[OVS_KEY_ATTR_ENCAP];
1091
1092	if (!is_mask)
1093		err = parse_flow_nlattrs(encap, a, key_attrs, log);
1094	else
1095		err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1096
1097	return err;
1098}
1099
1100static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1101				   u64 *key_attrs, const struct nlattr **a,
1102				   bool is_mask, bool log)
1103{
1104	int err;
1105	bool encap_valid = false;
1106
1107	err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1108					is_mask, log);
1109	if (err)
1110		return err;
1111
1112	encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_CFI_MASK));
1113	if (encap_valid) {
1114		err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1115						is_mask, log);
1116		if (err)
1117			return err;
1118	}
1119
1120	return 0;
1121}
1122
1123static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1124				       u64 *attrs, const struct nlattr **a,
1125				       bool is_mask, bool log)
1126{
1127	__be16 eth_type;
1128
1129	eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1130	if (is_mask) {
1131		/* Always exact match EtherType. */
1132		eth_type = htons(0xffff);
1133	} else if (!eth_proto_is_802_3(eth_type)) {
1134		OVS_NLERR(log, "EtherType %x is less than min %x",
1135				ntohs(eth_type), ETH_P_802_3_MIN);
1136		return -EINVAL;
1137	}
1138
1139	SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1140	*attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1141	return 0;
1142}
1143
1144static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1145				 u64 *attrs, const struct nlattr **a,
1146				 bool is_mask, bool log)
1147{
1148	u8 mac_proto = MAC_PROTO_ETHERNET;
1149
1150	if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1151		u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1152
1153		SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1154		*attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1155	}
1156
1157	if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1158		u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1159
1160		SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1161		*attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1162	}
1163
1164	if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1165		SW_FLOW_KEY_PUT(match, phy.priority,
1166			  nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1167		*attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1168	}
1169
1170	if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1171		u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1172
1173		if (is_mask) {
1174			in_port = 0xffffffff; /* Always exact match in_port. */
1175		} else if (in_port >= DP_MAX_PORTS) {
1176			OVS_NLERR(log, "Port %d exceeds max allowable %d",
1177				  in_port, DP_MAX_PORTS);
1178			return -EINVAL;
1179		}
1180
1181		SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1182		*attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1183	} else if (!is_mask) {
1184		SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1185	}
1186
1187	if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1188		uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1189
1190		SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1191		*attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1192	}
1193	if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1194		if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1195				       is_mask, log) < 0)
1196			return -EINVAL;
1197		*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1198	}
1199
1200	if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1201	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1202		u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1203
1204		if (ct_state & ~CT_SUPPORTED_MASK) {
1205			OVS_NLERR(log, "ct_state flags %08x unsupported",
1206				  ct_state);
1207			return -EINVAL;
1208		}
1209
1210		SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1211		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1212	}
1213	if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1214	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1215		u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1216
1217		SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1218		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1219	}
1220	if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1221	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1222		u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1223
1224		SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1225		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1226	}
1227	if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1228	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1229		const struct ovs_key_ct_labels *cl;
1230
1231		cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1232		SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1233				   sizeof(*cl), is_mask);
1234		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1235	}
1236	if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1237		const struct ovs_key_ct_tuple_ipv4 *ct;
1238
1239		ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1240
1241		SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1242		SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1243		SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1244		SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1245		SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1246		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1247	}
1248	if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1249		const struct ovs_key_ct_tuple_ipv6 *ct;
1250
1251		ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1252
1253		SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1254				   sizeof(match->key->ipv6.ct_orig.src),
1255				   is_mask);
1256		SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1257				   sizeof(match->key->ipv6.ct_orig.dst),
1258				   is_mask);
1259		SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1260		SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1261		SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1262		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1263	}
1264
1265	/* For layer 3 packets the Ethernet type is provided
1266	 * and treated as metadata but no MAC addresses are provided.
1267	 */
1268	if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1269	    (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1270		mac_proto = MAC_PROTO_NONE;
1271
1272	/* Always exact match mac_proto */
1273	SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1274
1275	if (mac_proto == MAC_PROTO_NONE)
1276		return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1277						   log);
1278
1279	return 0;
1280}
1281
1282int nsh_hdr_from_nlattr(const struct nlattr *attr,
1283			struct nshhdr *nh, size_t size)
1284{
1285	struct nlattr *a;
1286	int rem;
1287	u8 flags = 0;
1288	u8 ttl = 0;
1289	int mdlen = 0;
1290
1291	/* validate_nsh has check this, so we needn't do duplicate check here
1292	 */
1293	if (size < NSH_BASE_HDR_LEN)
1294		return -ENOBUFS;
1295
1296	nla_for_each_nested(a, attr, rem) {
1297		int type = nla_type(a);
1298
1299		switch (type) {
1300		case OVS_NSH_KEY_ATTR_BASE: {
1301			const struct ovs_nsh_key_base *base = nla_data(a);
1302
1303			flags = base->flags;
1304			ttl = base->ttl;
1305			nh->np = base->np;
1306			nh->mdtype = base->mdtype;
1307			nh->path_hdr = base->path_hdr;
1308			break;
1309		}
1310		case OVS_NSH_KEY_ATTR_MD1:
1311			mdlen = nla_len(a);
1312			if (mdlen > size - NSH_BASE_HDR_LEN)
1313				return -ENOBUFS;
1314			memcpy(&nh->md1, nla_data(a), mdlen);
1315			break;
1316
1317		case OVS_NSH_KEY_ATTR_MD2:
1318			mdlen = nla_len(a);
1319			if (mdlen > size - NSH_BASE_HDR_LEN)
1320				return -ENOBUFS;
1321			memcpy(&nh->md2, nla_data(a), mdlen);
1322			break;
1323
1324		default:
1325			return -EINVAL;
1326		}
1327	}
1328
1329	/* nsh header length  = NSH_BASE_HDR_LEN + mdlen */
1330	nh->ver_flags_ttl_len = 0;
1331	nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1332
1333	return 0;
1334}
1335
1336int nsh_key_from_nlattr(const struct nlattr *attr,
1337			struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1338{
1339	struct nlattr *a;
1340	int rem;
1341
1342	/* validate_nsh has check this, so we needn't do duplicate check here
1343	 */
1344	nla_for_each_nested(a, attr, rem) {
1345		int type = nla_type(a);
1346
1347		switch (type) {
1348		case OVS_NSH_KEY_ATTR_BASE: {
1349			const struct ovs_nsh_key_base *base = nla_data(a);
1350			const struct ovs_nsh_key_base *base_mask = base + 1;
1351
1352			nsh->base = *base;
1353			nsh_mask->base = *base_mask;
1354			break;
1355		}
1356		case OVS_NSH_KEY_ATTR_MD1: {
1357			const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1358			const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1359
1360			memcpy(nsh->context, md1->context, sizeof(*md1));
1361			memcpy(nsh_mask->context, md1_mask->context,
1362			       sizeof(*md1_mask));
1363			break;
1364		}
1365		case OVS_NSH_KEY_ATTR_MD2:
1366			/* Not supported yet */
1367			return -ENOTSUPP;
1368		default:
1369			return -EINVAL;
1370		}
1371	}
1372
1373	return 0;
1374}
1375
1376static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1377				   struct sw_flow_match *match, bool is_mask,
1378				   bool is_push_nsh, bool log)
1379{
1380	struct nlattr *a;
1381	int rem;
1382	bool has_base = false;
1383	bool has_md1 = false;
1384	bool has_md2 = false;
1385	u8 mdtype = 0;
1386	int mdlen = 0;
1387
1388	if (WARN_ON(is_push_nsh && is_mask))
1389		return -EINVAL;
1390
1391	nla_for_each_nested(a, attr, rem) {
1392		int type = nla_type(a);
1393		int i;
1394
1395		if (type > OVS_NSH_KEY_ATTR_MAX) {
1396			OVS_NLERR(log, "nsh attr %d is out of range max %d",
1397				  type, OVS_NSH_KEY_ATTR_MAX);
1398			return -EINVAL;
1399		}
1400
1401		if (!check_attr_len(nla_len(a),
1402				    ovs_nsh_key_attr_lens[type].len)) {
1403			OVS_NLERR(
1404			    log,
1405			    "nsh attr %d has unexpected len %d expected %d",
1406			    type,
1407			    nla_len(a),
1408			    ovs_nsh_key_attr_lens[type].len
1409			);
1410			return -EINVAL;
1411		}
1412
1413		switch (type) {
1414		case OVS_NSH_KEY_ATTR_BASE: {
1415			const struct ovs_nsh_key_base *base = nla_data(a);
1416
1417			has_base = true;
1418			mdtype = base->mdtype;
1419			SW_FLOW_KEY_PUT(match, nsh.base.flags,
1420					base->flags, is_mask);
1421			SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1422					base->ttl, is_mask);
1423			SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1424					base->mdtype, is_mask);
1425			SW_FLOW_KEY_PUT(match, nsh.base.np,
1426					base->np, is_mask);
1427			SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1428					base->path_hdr, is_mask);
1429			break;
1430		}
1431		case OVS_NSH_KEY_ATTR_MD1: {
1432			const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1433
1434			has_md1 = true;
1435			for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1436				SW_FLOW_KEY_PUT(match, nsh.context[i],
1437						md1->context[i], is_mask);
1438			break;
1439		}
1440		case OVS_NSH_KEY_ATTR_MD2:
1441			if (!is_push_nsh) /* Not supported MD type 2 yet */
1442				return -ENOTSUPP;
1443
1444			has_md2 = true;
1445			mdlen = nla_len(a);
1446			if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1447				OVS_NLERR(
1448				    log,
1449				    "Invalid MD length %d for MD type %d",
1450				    mdlen,
1451				    mdtype
1452				);
1453				return -EINVAL;
1454			}
1455			break;
1456		default:
1457			OVS_NLERR(log, "Unknown nsh attribute %d",
1458				  type);
1459			return -EINVAL;
1460		}
1461	}
1462
1463	if (rem > 0) {
1464		OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1465		return -EINVAL;
1466	}
1467
1468	if (has_md1 && has_md2) {
1469		OVS_NLERR(
1470		    1,
1471		    "invalid nsh attribute: md1 and md2 are exclusive."
1472		);
1473		return -EINVAL;
1474	}
1475
1476	if (!is_mask) {
1477		if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1478		    (has_md2 && mdtype != NSH_M_TYPE2)) {
1479			OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1480				  mdtype);
1481			return -EINVAL;
1482		}
1483
1484		if (is_push_nsh &&
1485		    (!has_base || (!has_md1 && !has_md2))) {
1486			OVS_NLERR(
1487			    1,
1488			    "push_nsh: missing base or metadata attributes"
1489			);
1490			return -EINVAL;
1491		}
1492	}
1493
1494	return 0;
1495}
1496
1497static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1498				u64 attrs, const struct nlattr **a,
1499				bool is_mask, bool log)
1500{
1501	int err;
1502
1503	err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1504	if (err)
1505		return err;
1506
1507	if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1508		const struct ovs_key_ethernet *eth_key;
1509
1510		eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1511		SW_FLOW_KEY_MEMCPY(match, eth.src,
1512				eth_key->eth_src, ETH_ALEN, is_mask);
1513		SW_FLOW_KEY_MEMCPY(match, eth.dst,
1514				eth_key->eth_dst, ETH_ALEN, is_mask);
1515		attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1516
1517		if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1518			/* VLAN attribute is always parsed before getting here since it
1519			 * may occur multiple times.
1520			 */
1521			OVS_NLERR(log, "VLAN attribute unexpected.");
1522			return -EINVAL;
1523		}
1524
1525		if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1526			err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1527							  log);
1528			if (err)
1529				return err;
1530		} else if (!is_mask) {
1531			SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1532		}
1533	} else if (!match->key->eth.type) {
1534		OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1535		return -EINVAL;
1536	}
1537
1538	if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1539		const struct ovs_key_ipv4 *ipv4_key;
1540
1541		ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1542		if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1543			OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1544				  ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1545			return -EINVAL;
1546		}
1547		SW_FLOW_KEY_PUT(match, ip.proto,
1548				ipv4_key->ipv4_proto, is_mask);
1549		SW_FLOW_KEY_PUT(match, ip.tos,
1550				ipv4_key->ipv4_tos, is_mask);
1551		SW_FLOW_KEY_PUT(match, ip.ttl,
1552				ipv4_key->ipv4_ttl, is_mask);
1553		SW_FLOW_KEY_PUT(match, ip.frag,
1554				ipv4_key->ipv4_frag, is_mask);
1555		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1556				ipv4_key->ipv4_src, is_mask);
1557		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1558				ipv4_key->ipv4_dst, is_mask);
1559		attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1560	}
1561
1562	if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1563		const struct ovs_key_ipv6 *ipv6_key;
1564
1565		ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1566		if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1567			OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1568				  ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1569			return -EINVAL;
1570		}
1571
1572		if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1573			OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1574				  ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1575			return -EINVAL;
1576		}
1577
1578		SW_FLOW_KEY_PUT(match, ipv6.label,
1579				ipv6_key->ipv6_label, is_mask);
1580		SW_FLOW_KEY_PUT(match, ip.proto,
1581				ipv6_key->ipv6_proto, is_mask);
1582		SW_FLOW_KEY_PUT(match, ip.tos,
1583				ipv6_key->ipv6_tclass, is_mask);
1584		SW_FLOW_KEY_PUT(match, ip.ttl,
1585				ipv6_key->ipv6_hlimit, is_mask);
1586		SW_FLOW_KEY_PUT(match, ip.frag,
1587				ipv6_key->ipv6_frag, is_mask);
1588		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1589				ipv6_key->ipv6_src,
1590				sizeof(match->key->ipv6.addr.src),
1591				is_mask);
1592		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1593				ipv6_key->ipv6_dst,
1594				sizeof(match->key->ipv6.addr.dst),
1595				is_mask);
1596
1597		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1598	}
1599
1600	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1601		const struct ovs_key_arp *arp_key;
1602
1603		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1604		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1605			OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1606				  arp_key->arp_op);
1607			return -EINVAL;
1608		}
1609
1610		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1611				arp_key->arp_sip, is_mask);
1612		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1613			arp_key->arp_tip, is_mask);
1614		SW_FLOW_KEY_PUT(match, ip.proto,
1615				ntohs(arp_key->arp_op), is_mask);
1616		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1617				arp_key->arp_sha, ETH_ALEN, is_mask);
1618		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1619				arp_key->arp_tha, ETH_ALEN, is_mask);
1620
1621		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1622	}
1623
1624	if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1625		if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1626					    is_mask, false, log) < 0)
1627			return -EINVAL;
1628		attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1629	}
1630
1631	if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1632		const struct ovs_key_mpls *mpls_key;
1633		u32 hdr_len;
1634		u32 label_count, label_count_mask, i;
1635
1636		mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1637		hdr_len = nla_len(a[OVS_KEY_ATTR_MPLS]);
1638		label_count = hdr_len / sizeof(struct ovs_key_mpls);
1639
1640		if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
1641		    hdr_len % sizeof(struct ovs_key_mpls))
1642			return -EINVAL;
1643
1644		label_count_mask =  GENMASK(label_count - 1, 0);
1645
1646		for (i = 0 ; i < label_count; i++)
1647			SW_FLOW_KEY_PUT(match, mpls.lse[i],
1648					mpls_key[i].mpls_lse, is_mask);
1649
1650		SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
1651				label_count_mask, is_mask);
1652
1653		attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1654	 }
1655
1656	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1657		const struct ovs_key_tcp *tcp_key;
1658
1659		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1660		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1661		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1662		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1663	}
1664
1665	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1666		SW_FLOW_KEY_PUT(match, tp.flags,
1667				nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1668				is_mask);
1669		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1670	}
1671
1672	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1673		const struct ovs_key_udp *udp_key;
1674
1675		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1676		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1677		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1678		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1679	}
1680
1681	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1682		const struct ovs_key_sctp *sctp_key;
1683
1684		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1685		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1686		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1687		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1688	}
1689
1690	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1691		const struct ovs_key_icmp *icmp_key;
1692
1693		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1694		SW_FLOW_KEY_PUT(match, tp.src,
1695				htons(icmp_key->icmp_type), is_mask);
1696		SW_FLOW_KEY_PUT(match, tp.dst,
1697				htons(icmp_key->icmp_code), is_mask);
1698		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1699	}
1700
1701	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1702		const struct ovs_key_icmpv6 *icmpv6_key;
1703
1704		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1705		SW_FLOW_KEY_PUT(match, tp.src,
1706				htons(icmpv6_key->icmpv6_type), is_mask);
1707		SW_FLOW_KEY_PUT(match, tp.dst,
1708				htons(icmpv6_key->icmpv6_code), is_mask);
1709		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1710	}
1711
1712	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1713		const struct ovs_key_nd *nd_key;
1714
1715		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1716		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1717			nd_key->nd_target,
1718			sizeof(match->key->ipv6.nd.target),
1719			is_mask);
1720		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1721			nd_key->nd_sll, ETH_ALEN, is_mask);
1722		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1723				nd_key->nd_tll, ETH_ALEN, is_mask);
1724		attrs &= ~(1 << OVS_KEY_ATTR_ND);
1725	}
1726
1727	if (attrs != 0) {
1728		OVS_NLERR(log, "Unknown key attributes %llx",
1729			  (unsigned long long)attrs);
1730		return -EINVAL;
1731	}
1732
1733	return 0;
1734}
1735
1736static void nlattr_set(struct nlattr *attr, u8 val,
1737		       const struct ovs_len_tbl *tbl)
1738{
1739	struct nlattr *nla;
1740	int rem;
1741
1742	/* The nlattr stream should already have been validated */
1743	nla_for_each_nested(nla, attr, rem) {
1744		if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1745			nlattr_set(nla, val, tbl[nla_type(nla)].next ? : tbl);
1746		else
1747			memset(nla_data(nla), val, nla_len(nla));
1748
1749		if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1750			*(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1751	}
1752}
1753
1754static void mask_set_nlattr(struct nlattr *attr, u8 val)
1755{
1756	nlattr_set(attr, val, ovs_key_lens);
1757}
1758
1759/**
1760 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1761 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1762 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1763 * does not include any don't care bit.
1764 * @net: Used to determine per-namespace field support.
1765 * @match: receives the extracted flow match information.
1766 * @nla_key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1767 * sequence. The fields should of the packet that triggered the creation
1768 * of this flow.
1769 * @nla_mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_*
1770 * Netlink attribute specifies the mask field of the wildcarded flow.
1771 * @log: Boolean to allow kernel error logging.  Normally true, but when
1772 * probing for feature compatibility this should be passed in as false to
1773 * suppress unnecessary error logging.
1774 */
1775int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1776		      const struct nlattr *nla_key,
1777		      const struct nlattr *nla_mask,
1778		      bool log)
1779{
1780	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1781	struct nlattr *newmask = NULL;
1782	u64 key_attrs = 0;
1783	u64 mask_attrs = 0;
1784	int err;
1785
1786	err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1787	if (err)
1788		return err;
1789
1790	err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1791	if (err)
1792		return err;
1793
1794	err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1795	if (err)
1796		return err;
1797
1798	if (match->mask) {
1799		if (!nla_mask) {
1800			/* Create an exact match mask. We need to set to 0xff
1801			 * all the 'match->mask' fields that have been touched
1802			 * in 'match->key'. We cannot simply memset
1803			 * 'match->mask', because padding bytes and fields not
1804			 * specified in 'match->key' should be left to 0.
1805			 * Instead, we use a stream of netlink attributes,
1806			 * copied from 'key' and set to 0xff.
1807			 * ovs_key_from_nlattrs() will take care of filling
1808			 * 'match->mask' appropriately.
1809			 */
1810			newmask = kmemdup(nla_key,
1811					  nla_total_size(nla_len(nla_key)),
1812					  GFP_KERNEL);
1813			if (!newmask)
1814				return -ENOMEM;
1815
1816			mask_set_nlattr(newmask, 0xff);
1817
1818			/* The userspace does not send tunnel attributes that
1819			 * are 0, but we should not wildcard them nonetheless.
1820			 */
1821			if (match->key->tun_proto)
1822				SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1823							 0xff, true);
1824
1825			nla_mask = newmask;
1826		}
1827
1828		err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1829		if (err)
1830			goto free_newmask;
1831
1832		/* Always match on tci. */
1833		SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1834		SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1835
1836		err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1837		if (err)
1838			goto free_newmask;
1839
1840		err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1841					   log);
1842		if (err)
1843			goto free_newmask;
1844	}
1845
1846	if (!match_validate(match, key_attrs, mask_attrs, log))
1847		err = -EINVAL;
1848
1849free_newmask:
1850	kfree(newmask);
1851	return err;
1852}
1853
1854static size_t get_ufid_len(const struct nlattr *attr, bool log)
1855{
1856	size_t len;
1857
1858	if (!attr)
1859		return 0;
1860
1861	len = nla_len(attr);
1862	if (len < 1 || len > MAX_UFID_LENGTH) {
1863		OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1864			  nla_len(attr), MAX_UFID_LENGTH);
1865		return 0;
1866	}
1867
1868	return len;
1869}
1870
1871/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1872 * or false otherwise.
1873 */
1874bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1875		      bool log)
1876{
1877	sfid->ufid_len = get_ufid_len(attr, log);
1878	if (sfid->ufid_len)
1879		memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1880
1881	return sfid->ufid_len;
1882}
1883
1884int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1885			   const struct sw_flow_key *key, bool log)
1886{
1887	struct sw_flow_key *new_key;
1888
1889	if (ovs_nla_get_ufid(sfid, ufid, log))
1890		return 0;
1891
1892	/* If UFID was not provided, use unmasked key. */
1893	new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1894	if (!new_key)
1895		return -ENOMEM;
1896	memcpy(new_key, key, sizeof(*key));
1897	sfid->unmasked_key = new_key;
1898
1899	return 0;
1900}
1901
1902u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1903{
1904	return attr ? nla_get_u32(attr) : 0;
1905}
1906
1907/**
1908 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1909 * @net: Network namespace.
1910 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1911 * metadata.
1912 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1913 * attributes.
1914 * @attrs: Bit mask for the netlink attributes included in @a.
1915 * @log: Boolean to allow kernel error logging.  Normally true, but when
1916 * probing for feature compatibility this should be passed in as false to
1917 * suppress unnecessary error logging.
1918 *
1919 * This parses a series of Netlink attributes that form a flow key, which must
1920 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1921 * get the metadata, that is, the parts of the flow key that cannot be
1922 * extracted from the packet itself.
1923 *
1924 * This must be called before the packet key fields are filled in 'key'.
1925 */
1926
1927int ovs_nla_get_flow_metadata(struct net *net,
1928			      const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1929			      u64 attrs, struct sw_flow_key *key, bool log)
1930{
1931	struct sw_flow_match match;
1932
1933	memset(&match, 0, sizeof(match));
1934	match.key = key;
1935
1936	key->ct_state = 0;
1937	key->ct_zone = 0;
1938	key->ct_orig_proto = 0;
1939	memset(&key->ct, 0, sizeof(key->ct));
1940	memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1941	memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1942
1943	key->phy.in_port = DP_MAX_PORTS;
1944
1945	return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1946}
1947
1948static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1949			    bool is_mask)
1950{
1951	__be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1952
1953	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1954	    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1955		return -EMSGSIZE;
1956	return 0;
1957}
1958
1959static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1960			     struct sk_buff *skb)
1961{
1962	struct nlattr *start;
1963
1964	start = nla_nest_start_noflag(skb, OVS_KEY_ATTR_NSH);
1965	if (!start)
1966		return -EMSGSIZE;
1967
1968	if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
1969		goto nla_put_failure;
1970
1971	if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
1972		if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
1973			    sizeof(nsh->context), nsh->context))
1974			goto nla_put_failure;
1975	}
1976
1977	/* Don't support MD type 2 yet */
1978
1979	nla_nest_end(skb, start);
1980
1981	return 0;
1982
1983nla_put_failure:
1984	return -EMSGSIZE;
1985}
1986
1987static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1988			     const struct sw_flow_key *output, bool is_mask,
1989			     struct sk_buff *skb)
1990{
1991	struct ovs_key_ethernet *eth_key;
1992	struct nlattr *nla;
1993	struct nlattr *encap = NULL;
1994	struct nlattr *in_encap = NULL;
1995
1996	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1997		goto nla_put_failure;
1998
1999	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
2000		goto nla_put_failure;
2001
2002	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
2003		goto nla_put_failure;
2004
2005	if ((swkey->tun_proto || is_mask)) {
2006		const void *opts = NULL;
2007
2008		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
2009			opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
2010
2011		if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
2012				     swkey->tun_opts_len, swkey->tun_proto, 0))
2013			goto nla_put_failure;
2014	}
2015
2016	if (swkey->phy.in_port == DP_MAX_PORTS) {
2017		if (is_mask && (output->phy.in_port == 0xffff))
2018			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
2019				goto nla_put_failure;
2020	} else {
2021		u16 upper_u16;
2022		upper_u16 = !is_mask ? 0 : 0xffff;
2023
2024		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
2025				(upper_u16 << 16) | output->phy.in_port))
2026			goto nla_put_failure;
2027	}
2028
2029	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2030		goto nla_put_failure;
2031
2032	if (ovs_ct_put_key(swkey, output, skb))
2033		goto nla_put_failure;
2034
2035	if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2036		nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2037		if (!nla)
2038			goto nla_put_failure;
2039
2040		eth_key = nla_data(nla);
2041		ether_addr_copy(eth_key->eth_src, output->eth.src);
2042		ether_addr_copy(eth_key->eth_dst, output->eth.dst);
2043
2044		if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
2045			if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
2046				goto nla_put_failure;
2047			encap = nla_nest_start_noflag(skb, OVS_KEY_ATTR_ENCAP);
2048			if (!swkey->eth.vlan.tci)
2049				goto unencap;
2050
2051			if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
2052				if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
2053					goto nla_put_failure;
2054				in_encap = nla_nest_start_noflag(skb,
2055								 OVS_KEY_ATTR_ENCAP);
2056				if (!swkey->eth.cvlan.tci)
2057					goto unencap;
2058			}
2059		}
2060
2061		if (swkey->eth.type == htons(ETH_P_802_2)) {
2062			/*
2063			* Ethertype 802.2 is represented in the netlink with omitted
2064			* OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2065			* 0xffff in the mask attribute.  Ethertype can also
2066			* be wildcarded.
2067			*/
2068			if (is_mask && output->eth.type)
2069				if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2070							output->eth.type))
2071					goto nla_put_failure;
2072			goto unencap;
2073		}
2074	}
2075
2076	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2077		goto nla_put_failure;
2078
2079	if (eth_type_vlan(swkey->eth.type)) {
2080		/* There are 3 VLAN tags, we don't know anything about the rest
2081		 * of the packet, so truncate here.
2082		 */
2083		WARN_ON_ONCE(!(encap && in_encap));
2084		goto unencap;
2085	}
2086
2087	if (swkey->eth.type == htons(ETH_P_IP)) {
2088		struct ovs_key_ipv4 *ipv4_key;
2089
2090		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2091		if (!nla)
2092			goto nla_put_failure;
2093		ipv4_key = nla_data(nla);
2094		ipv4_key->ipv4_src = output->ipv4.addr.src;
2095		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2096		ipv4_key->ipv4_proto = output->ip.proto;
2097		ipv4_key->ipv4_tos = output->ip.tos;
2098		ipv4_key->ipv4_ttl = output->ip.ttl;
2099		ipv4_key->ipv4_frag = output->ip.frag;
2100	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2101		struct ovs_key_ipv6 *ipv6_key;
2102
2103		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2104		if (!nla)
2105			goto nla_put_failure;
2106		ipv6_key = nla_data(nla);
2107		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2108				sizeof(ipv6_key->ipv6_src));
2109		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2110				sizeof(ipv6_key->ipv6_dst));
2111		ipv6_key->ipv6_label = output->ipv6.label;
2112		ipv6_key->ipv6_proto = output->ip.proto;
2113		ipv6_key->ipv6_tclass = output->ip.tos;
2114		ipv6_key->ipv6_hlimit = output->ip.ttl;
2115		ipv6_key->ipv6_frag = output->ip.frag;
2116	} else if (swkey->eth.type == htons(ETH_P_NSH)) {
2117		if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2118			goto nla_put_failure;
2119	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
2120		   swkey->eth.type == htons(ETH_P_RARP)) {
2121		struct ovs_key_arp *arp_key;
2122
2123		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2124		if (!nla)
2125			goto nla_put_failure;
2126		arp_key = nla_data(nla);
2127		memset(arp_key, 0, sizeof(struct ovs_key_arp));
2128		arp_key->arp_sip = output->ipv4.addr.src;
2129		arp_key->arp_tip = output->ipv4.addr.dst;
2130		arp_key->arp_op = htons(output->ip.proto);
2131		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2132		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2133	} else if (eth_p_mpls(swkey->eth.type)) {
2134		u8 i, num_labels;
2135		struct ovs_key_mpls *mpls_key;
2136
2137		num_labels = hweight_long(output->mpls.num_labels_mask);
2138		nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS,
2139				  num_labels * sizeof(*mpls_key));
2140		if (!nla)
2141			goto nla_put_failure;
2142
2143		mpls_key = nla_data(nla);
2144		for (i = 0; i < num_labels; i++)
2145			mpls_key[i].mpls_lse = output->mpls.lse[i];
2146	}
2147
2148	if ((swkey->eth.type == htons(ETH_P_IP) ||
2149	     swkey->eth.type == htons(ETH_P_IPV6)) &&
2150	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2151
2152		if (swkey->ip.proto == IPPROTO_TCP) {
2153			struct ovs_key_tcp *tcp_key;
2154
2155			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2156			if (!nla)
2157				goto nla_put_failure;
2158			tcp_key = nla_data(nla);
2159			tcp_key->tcp_src = output->tp.src;
2160			tcp_key->tcp_dst = output->tp.dst;
2161			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2162					 output->tp.flags))
2163				goto nla_put_failure;
2164		} else if (swkey->ip.proto == IPPROTO_UDP) {
2165			struct ovs_key_udp *udp_key;
2166
2167			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2168			if (!nla)
2169				goto nla_put_failure;
2170			udp_key = nla_data(nla);
2171			udp_key->udp_src = output->tp.src;
2172			udp_key->udp_dst = output->tp.dst;
2173		} else if (swkey->ip.proto == IPPROTO_SCTP) {
2174			struct ovs_key_sctp *sctp_key;
2175
2176			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2177			if (!nla)
2178				goto nla_put_failure;
2179			sctp_key = nla_data(nla);
2180			sctp_key->sctp_src = output->tp.src;
2181			sctp_key->sctp_dst = output->tp.dst;
2182		} else if (swkey->eth.type == htons(ETH_P_IP) &&
2183			   swkey->ip.proto == IPPROTO_ICMP) {
2184			struct ovs_key_icmp *icmp_key;
2185
2186			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2187			if (!nla)
2188				goto nla_put_failure;
2189			icmp_key = nla_data(nla);
2190			icmp_key->icmp_type = ntohs(output->tp.src);
2191			icmp_key->icmp_code = ntohs(output->tp.dst);
2192		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2193			   swkey->ip.proto == IPPROTO_ICMPV6) {
2194			struct ovs_key_icmpv6 *icmpv6_key;
2195
2196			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2197						sizeof(*icmpv6_key));
2198			if (!nla)
2199				goto nla_put_failure;
2200			icmpv6_key = nla_data(nla);
2201			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2202			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2203
2204			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
2205			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
2206				struct ovs_key_nd *nd_key;
2207
2208				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2209				if (!nla)
2210					goto nla_put_failure;
2211				nd_key = nla_data(nla);
2212				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2213							sizeof(nd_key->nd_target));
2214				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2215				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2216			}
2217		}
2218	}
2219
2220unencap:
2221	if (in_encap)
2222		nla_nest_end(skb, in_encap);
2223	if (encap)
2224		nla_nest_end(skb, encap);
2225
2226	return 0;
2227
2228nla_put_failure:
2229	return -EMSGSIZE;
2230}
2231
2232int ovs_nla_put_key(const struct sw_flow_key *swkey,
2233		    const struct sw_flow_key *output, int attr, bool is_mask,
2234		    struct sk_buff *skb)
2235{
2236	int err;
2237	struct nlattr *nla;
2238
2239	nla = nla_nest_start_noflag(skb, attr);
2240	if (!nla)
2241		return -EMSGSIZE;
2242	err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2243	if (err)
2244		return err;
2245	nla_nest_end(skb, nla);
2246
2247	return 0;
2248}
2249
2250/* Called with ovs_mutex or RCU read lock. */
2251int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2252{
2253	if (ovs_identifier_is_ufid(&flow->id))
2254		return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2255			       flow->id.ufid);
2256
2257	return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2258			       OVS_FLOW_ATTR_KEY, false, skb);
2259}
2260
2261/* Called with ovs_mutex or RCU read lock. */
2262int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2263{
2264	return ovs_nla_put_key(&flow->key, &flow->key,
2265				OVS_FLOW_ATTR_KEY, false, skb);
2266}
2267
2268/* Called with ovs_mutex or RCU read lock. */
2269int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2270{
2271	return ovs_nla_put_key(&flow->key, &flow->mask->key,
2272				OVS_FLOW_ATTR_MASK, true, skb);
2273}
2274
2275#define MAX_ACTIONS_BUFSIZE	(32 * 1024)
2276
2277static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2278{
2279	struct sw_flow_actions *sfa;
2280
2281	WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2282
2283	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
2284	if (!sfa)
2285		return ERR_PTR(-ENOMEM);
2286
2287	sfa->actions_len = 0;
2288	return sfa;
2289}
2290
2291static void ovs_nla_free_set_action(const struct nlattr *a)
2292{
2293	const struct nlattr *ovs_key = nla_data(a);
2294	struct ovs_tunnel_info *ovs_tun;
2295
2296	switch (nla_type(ovs_key)) {
2297	case OVS_KEY_ATTR_TUNNEL_INFO:
2298		ovs_tun = nla_data(ovs_key);
2299		dst_release((struct dst_entry *)ovs_tun->tun_dst);
2300		break;
2301	}
2302}
2303
2304void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2305{
2306	const struct nlattr *a;
2307	int rem;
2308
2309	if (!sf_acts)
2310		return;
2311
2312	nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
2313		switch (nla_type(a)) {
2314		case OVS_ACTION_ATTR_SET:
2315			ovs_nla_free_set_action(a);
2316			break;
2317		case OVS_ACTION_ATTR_CT:
2318			ovs_ct_free_action(a);
2319			break;
2320		}
2321	}
2322
2323	kfree(sf_acts);
2324}
2325
2326static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2327{
2328	ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2329}
2330
2331/* Schedules 'sf_acts' to be freed after the next RCU grace period.
2332 * The caller must hold rcu_read_lock for this to be sensible. */
2333void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2334{
2335	call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2336}
2337
2338static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2339				       int attr_len, bool log)
2340{
2341
2342	struct sw_flow_actions *acts;
2343	int new_acts_size;
2344	size_t req_size = NLA_ALIGN(attr_len);
2345	int next_offset = offsetof(struct sw_flow_actions, actions) +
2346					(*sfa)->actions_len;
2347
2348	if (req_size <= (ksize(*sfa) - next_offset))
2349		goto out;
2350
2351	new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2352
2353	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2354		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size) {
2355			OVS_NLERR(log, "Flow action size exceeds max %u",
2356				  MAX_ACTIONS_BUFSIZE);
2357			return ERR_PTR(-EMSGSIZE);
2358		}
2359		new_acts_size = MAX_ACTIONS_BUFSIZE;
2360	}
2361
2362	acts = nla_alloc_flow_actions(new_acts_size);
2363	if (IS_ERR(acts))
2364		return (void *)acts;
2365
2366	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2367	acts->actions_len = (*sfa)->actions_len;
2368	acts->orig_len = (*sfa)->orig_len;
2369	kfree(*sfa);
2370	*sfa = acts;
2371
2372out:
2373	(*sfa)->actions_len += req_size;
2374	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2375}
2376
2377static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2378				   int attrtype, void *data, int len, bool log)
2379{
2380	struct nlattr *a;
2381
2382	a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2383	if (IS_ERR(a))
2384		return a;
2385
2386	a->nla_type = attrtype;
2387	a->nla_len = nla_attr_size(len);
2388
2389	if (data)
2390		memcpy(nla_data(a), data, len);
2391	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2392
2393	return a;
2394}
2395
2396int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2397		       int len, bool log)
2398{
2399	struct nlattr *a;
2400
2401	a = __add_action(sfa, attrtype, data, len, log);
2402
2403	return PTR_ERR_OR_ZERO(a);
2404}
2405
2406static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2407					  int attrtype, bool log)
2408{
2409	int used = (*sfa)->actions_len;
2410	int err;
2411
2412	err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2413	if (err)
2414		return err;
2415
2416	return used;
2417}
2418
2419static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2420					 int st_offset)
2421{
2422	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2423							       st_offset);
2424
2425	a->nla_len = sfa->actions_len - st_offset;
2426}
2427
2428static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2429				  const struct sw_flow_key *key,
2430				  struct sw_flow_actions **sfa,
2431				  __be16 eth_type, __be16 vlan_tci,
2432				  u32 mpls_label_count, bool log);
2433
2434static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2435				    const struct sw_flow_key *key,
2436				    struct sw_flow_actions **sfa,
2437				    __be16 eth_type, __be16 vlan_tci,
2438				    u32 mpls_label_count, bool log, bool last)
2439{
2440	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2441	const struct nlattr *probability, *actions;
2442	const struct nlattr *a;
2443	int rem, start, err;
2444	struct sample_arg arg;
2445
2446	memset(attrs, 0, sizeof(attrs));
2447	nla_for_each_nested(a, attr, rem) {
2448		int type = nla_type(a);
2449		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2450			return -EINVAL;
2451		attrs[type] = a;
2452	}
2453	if (rem)
2454		return -EINVAL;
2455
2456	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2457	if (!probability || nla_len(probability) != sizeof(u32))
2458		return -EINVAL;
2459
2460	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2461	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2462		return -EINVAL;
2463
2464	/* validation done, copy sample action. */
2465	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2466	if (start < 0)
2467		return start;
2468
2469	/* When both skb and flow may be changed, put the sample
2470	 * into a deferred fifo. On the other hand, if only skb
2471	 * may be modified, the actions can be executed in place.
2472	 *
2473	 * Do this analysis at the flow installation time.
2474	 * Set 'clone_action->exec' to true if the actions can be
2475	 * executed without being deferred.
2476	 *
2477	 * If the sample is the last action, it can always be excuted
2478	 * rather than deferred.
2479	 */
2480	arg.exec = last || !actions_may_change_flow(actions);
2481	arg.probability = nla_get_u32(probability);
2482
2483	err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2484				 log);
2485	if (err)
2486		return err;
2487
2488	err = __ovs_nla_copy_actions(net, actions, key, sfa,
2489				     eth_type, vlan_tci, mpls_label_count, log);
2490
2491	if (err)
2492		return err;
2493
2494	add_nested_action_end(*sfa, start);
2495
2496	return 0;
2497}
2498
2499static int validate_and_copy_dec_ttl(struct net *net,
2500				     const struct nlattr *attr,
2501				     const struct sw_flow_key *key,
2502				     struct sw_flow_actions **sfa,
2503				     __be16 eth_type, __be16 vlan_tci,
2504				     u32 mpls_label_count, bool log)
2505{
2506	const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2507	int start, action_start, err, rem;
2508	const struct nlattr *a, *actions;
2509
2510	memset(attrs, 0, sizeof(attrs));
2511	nla_for_each_nested(a, attr, rem) {
2512		int type = nla_type(a);
2513
2514		/* Ignore unknown attributes to be future proof. */
2515		if (type > OVS_DEC_TTL_ATTR_MAX)
2516			continue;
2517
2518		if (!type || attrs[type]) {
2519			OVS_NLERR(log, "Duplicate or invalid key (type %d).",
2520				  type);
2521			return -EINVAL;
2522		}
2523
2524		attrs[type] = a;
2525	}
2526
2527	if (rem) {
2528		OVS_NLERR(log, "Message has %d unknown bytes.", rem);
2529		return -EINVAL;
2530	}
2531
2532	actions = attrs[OVS_DEC_TTL_ATTR_ACTION];
2533	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN)) {
2534		OVS_NLERR(log, "Missing valid actions attribute.");
2535		return -EINVAL;
2536	}
2537
2538	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2539	if (start < 0)
2540		return start;
2541
2542	action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2543	if (action_start < 0)
2544		return action_start;
2545
2546	err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2547				     vlan_tci, mpls_label_count, log);
2548	if (err)
2549		return err;
2550
2551	add_nested_action_end(*sfa, action_start);
2552	add_nested_action_end(*sfa, start);
2553	return 0;
2554}
2555
2556static int validate_and_copy_clone(struct net *net,
2557				   const struct nlattr *attr,
2558				   const struct sw_flow_key *key,
2559				   struct sw_flow_actions **sfa,
2560				   __be16 eth_type, __be16 vlan_tci,
2561				   u32 mpls_label_count, bool log, bool last)
2562{
2563	int start, err;
2564	u32 exec;
2565
2566	if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2567		return -EINVAL;
2568
2569	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2570	if (start < 0)
2571		return start;
2572
2573	exec = last || !actions_may_change_flow(attr);
2574
2575	err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2576				 sizeof(exec), log);
2577	if (err)
2578		return err;
2579
2580	err = __ovs_nla_copy_actions(net, attr, key, sfa,
2581				     eth_type, vlan_tci, mpls_label_count, log);
2582	if (err)
2583		return err;
2584
2585	add_nested_action_end(*sfa, start);
2586
2587	return 0;
2588}
2589
2590void ovs_match_init(struct sw_flow_match *match,
2591		    struct sw_flow_key *key,
2592		    bool reset_key,
2593		    struct sw_flow_mask *mask)
2594{
2595	memset(match, 0, sizeof(*match));
2596	match->key = key;
2597	match->mask = mask;
2598
2599	if (reset_key)
2600		memset(key, 0, sizeof(*key));
2601
2602	if (mask) {
2603		memset(&mask->key, 0, sizeof(mask->key));
2604		mask->range.start = mask->range.end = 0;
2605	}
2606}
2607
2608static int validate_geneve_opts(struct sw_flow_key *key)
2609{
2610	struct geneve_opt *option;
2611	int opts_len = key->tun_opts_len;
2612	bool crit_opt = false;
2613
2614	option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2615	while (opts_len > 0) {
2616		int len;
2617
2618		if (opts_len < sizeof(*option))
2619			return -EINVAL;
2620
2621		len = sizeof(*option) + option->length * 4;
2622		if (len > opts_len)
2623			return -EINVAL;
2624
2625		crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2626
2627		option = (struct geneve_opt *)((u8 *)option + len);
2628		opts_len -= len;
2629	}
2630
2631	key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2632
2633	return 0;
2634}
2635
2636static int validate_and_copy_set_tun(const struct nlattr *attr,
2637				     struct sw_flow_actions **sfa, bool log)
2638{
2639	struct sw_flow_match match;
2640	struct sw_flow_key key;
2641	struct metadata_dst *tun_dst;
2642	struct ip_tunnel_info *tun_info;
2643	struct ovs_tunnel_info *ovs_tun;
2644	struct nlattr *a;
2645	int err = 0, start, opts_type;
2646	__be16 dst_opt_type;
2647
2648	dst_opt_type = 0;
2649	ovs_match_init(&match, &key, true, NULL);
2650	opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2651	if (opts_type < 0)
2652		return opts_type;
2653
2654	if (key.tun_opts_len) {
2655		switch (opts_type) {
2656		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2657			err = validate_geneve_opts(&key);
2658			if (err < 0)
2659				return err;
2660			dst_opt_type = TUNNEL_GENEVE_OPT;
2661			break;
2662		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2663			dst_opt_type = TUNNEL_VXLAN_OPT;
2664			break;
2665		case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2666			dst_opt_type = TUNNEL_ERSPAN_OPT;
2667			break;
2668		}
2669	}
2670
2671	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2672	if (start < 0)
2673		return start;
2674
2675	tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2676				     GFP_KERNEL);
2677
2678	if (!tun_dst)
2679		return -ENOMEM;
2680
2681	err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2682	if (err) {
2683		dst_release((struct dst_entry *)tun_dst);
2684		return err;
2685	}
2686
2687	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2688			 sizeof(*ovs_tun), log);
2689	if (IS_ERR(a)) {
2690		dst_release((struct dst_entry *)tun_dst);
2691		return PTR_ERR(a);
2692	}
2693
2694	ovs_tun = nla_data(a);
2695	ovs_tun->tun_dst = tun_dst;
2696
2697	tun_info = &tun_dst->u.tun_info;
2698	tun_info->mode = IP_TUNNEL_INFO_TX;
2699	if (key.tun_proto == AF_INET6)
2700		tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2701	else if (key.tun_proto == AF_INET && key.tun_key.u.ipv4.dst == 0)
2702		tun_info->mode |= IP_TUNNEL_INFO_BRIDGE;
2703	tun_info->key = key.tun_key;
2704
2705	/* We need to store the options in the action itself since
2706	 * everything else will go away after flow setup. We can append
2707	 * it to tun_info and then point there.
2708	 */
2709	ip_tunnel_info_opts_set(tun_info,
2710				TUN_METADATA_OPTS(&key, key.tun_opts_len),
2711				key.tun_opts_len, dst_opt_type);
2712	add_nested_action_end(*sfa, start);
2713
2714	return err;
2715}
2716
2717static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2718			 bool is_push_nsh, bool log)
2719{
2720	struct sw_flow_match match;
2721	struct sw_flow_key key;
2722	int ret = 0;
2723
2724	ovs_match_init(&match, &key, true, NULL);
2725	ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2726				      is_push_nsh, log);
2727	return !ret;
2728}
2729
2730/* Return false if there are any non-masked bits set.
2731 * Mask follows data immediately, before any netlink padding.
2732 */
2733static bool validate_masked(u8 *data, int len)
2734{
2735	u8 *mask = data + len;
2736
2737	while (len--)
2738		if (*data++ & ~*mask++)
2739			return false;
2740
2741	return true;
2742}
2743
2744static int validate_set(const struct nlattr *a,
2745			const struct sw_flow_key *flow_key,
2746			struct sw_flow_actions **sfa, bool *skip_copy,
2747			u8 mac_proto, __be16 eth_type, bool masked, bool log)
2748{
2749	const struct nlattr *ovs_key = nla_data(a);
2750	int key_type = nla_type(ovs_key);
2751	size_t key_len;
2752
2753	/* There can be only one key in a action */
2754	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2755		return -EINVAL;
2756
2757	key_len = nla_len(ovs_key);
2758	if (masked)
2759		key_len /= 2;
2760
2761	if (key_type > OVS_KEY_ATTR_MAX ||
2762	    !check_attr_len(key_len, ovs_key_lens[key_type].len))
2763		return -EINVAL;
2764
2765	if (masked && !validate_masked(nla_data(ovs_key), key_len))
2766		return -EINVAL;
2767
2768	switch (key_type) {
 
 
 
 
2769	case OVS_KEY_ATTR_PRIORITY:
2770	case OVS_KEY_ATTR_SKB_MARK:
2771	case OVS_KEY_ATTR_CT_MARK:
2772	case OVS_KEY_ATTR_CT_LABELS:
2773		break;
2774
2775	case OVS_KEY_ATTR_ETHERNET:
2776		if (mac_proto != MAC_PROTO_ETHERNET)
2777			return -EINVAL;
2778		break;
2779
2780	case OVS_KEY_ATTR_TUNNEL: {
2781		int err;
2782
2783		if (masked)
2784			return -EINVAL; /* Masked tunnel set not supported. */
2785
2786		*skip_copy = true;
2787		err = validate_and_copy_set_tun(a, sfa, log);
2788		if (err)
2789			return err;
2790		break;
2791	}
2792	case OVS_KEY_ATTR_IPV4: {
2793		const struct ovs_key_ipv4 *ipv4_key;
2794
 
2795		if (eth_type != htons(ETH_P_IP))
2796			return -EINVAL;
2797
2798		ipv4_key = nla_data(ovs_key);
2799
2800		if (masked) {
2801			const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2802
2803			/* Non-writeable fields. */
2804			if (mask->ipv4_proto || mask->ipv4_frag)
2805				return -EINVAL;
2806		} else {
2807			if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2808				return -EINVAL;
2809
2810			if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2811				return -EINVAL;
2812		}
2813		break;
2814	}
2815	case OVS_KEY_ATTR_IPV6: {
2816		const struct ovs_key_ipv6 *ipv6_key;
2817
 
2818		if (eth_type != htons(ETH_P_IPV6))
2819			return -EINVAL;
2820
2821		ipv6_key = nla_data(ovs_key);
2822
2823		if (masked) {
2824			const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2825
2826			/* Non-writeable fields. */
2827			if (mask->ipv6_proto || mask->ipv6_frag)
2828				return -EINVAL;
2829
2830			/* Invalid bits in the flow label mask? */
2831			if (ntohl(mask->ipv6_label) & 0xFFF00000)
2832				return -EINVAL;
2833		} else {
2834			if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2835				return -EINVAL;
2836
2837			if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2838				return -EINVAL;
2839		}
2840		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2841			return -EINVAL;
2842
2843		break;
2844	}
2845	case OVS_KEY_ATTR_TCP:
2846		if ((eth_type != htons(ETH_P_IP) &&
2847		     eth_type != htons(ETH_P_IPV6)) ||
2848		    flow_key->ip.proto != IPPROTO_TCP)
2849			return -EINVAL;
2850
2851		break;
2852
2853	case OVS_KEY_ATTR_UDP:
2854		if ((eth_type != htons(ETH_P_IP) &&
2855		     eth_type != htons(ETH_P_IPV6)) ||
2856		    flow_key->ip.proto != IPPROTO_UDP)
2857			return -EINVAL;
2858
2859		break;
2860
2861	case OVS_KEY_ATTR_MPLS:
2862		if (!eth_p_mpls(eth_type))
2863			return -EINVAL;
2864		break;
2865
2866	case OVS_KEY_ATTR_SCTP:
2867		if ((eth_type != htons(ETH_P_IP) &&
2868		     eth_type != htons(ETH_P_IPV6)) ||
2869		    flow_key->ip.proto != IPPROTO_SCTP)
2870			return -EINVAL;
2871
2872		break;
2873
2874	case OVS_KEY_ATTR_NSH:
2875		if (eth_type != htons(ETH_P_NSH))
2876			return -EINVAL;
2877		if (!validate_nsh(nla_data(a), masked, false, log))
2878			return -EINVAL;
2879		break;
2880
2881	default:
2882		return -EINVAL;
2883	}
2884
2885	/* Convert non-masked non-tunnel set actions to masked set actions. */
2886	if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2887		int start, len = key_len * 2;
2888		struct nlattr *at;
2889
2890		*skip_copy = true;
2891
2892		start = add_nested_action_start(sfa,
2893						OVS_ACTION_ATTR_SET_TO_MASKED,
2894						log);
2895		if (start < 0)
2896			return start;
2897
2898		at = __add_action(sfa, key_type, NULL, len, log);
2899		if (IS_ERR(at))
2900			return PTR_ERR(at);
2901
2902		memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2903		memset(nla_data(at) + key_len, 0xff, key_len);    /* Mask. */
2904		/* Clear non-writeable bits from otherwise writeable fields. */
2905		if (key_type == OVS_KEY_ATTR_IPV6) {
2906			struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2907
2908			mask->ipv6_label &= htonl(0x000FFFFF);
2909		}
2910		add_nested_action_end(*sfa, start);
2911	}
2912
2913	return 0;
2914}
2915
2916static int validate_userspace(const struct nlattr *attr)
2917{
2918	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2919		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2920		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2921		[OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2922	};
2923	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2924	int error;
2925
2926	error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
2927					    userspace_policy, NULL);
2928	if (error)
2929		return error;
2930
2931	if (!a[OVS_USERSPACE_ATTR_PID] ||
2932	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2933		return -EINVAL;
2934
2935	return 0;
2936}
2937
2938static const struct nla_policy cpl_policy[OVS_CHECK_PKT_LEN_ATTR_MAX + 1] = {
2939	[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] = {.type = NLA_U16 },
2940	[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER] = {.type = NLA_NESTED },
2941	[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL] = {.type = NLA_NESTED },
2942};
2943
2944static int validate_and_copy_check_pkt_len(struct net *net,
2945					   const struct nlattr *attr,
2946					   const struct sw_flow_key *key,
2947					   struct sw_flow_actions **sfa,
2948					   __be16 eth_type, __be16 vlan_tci,
2949					   u32 mpls_label_count,
2950					   bool log, bool last)
2951{
2952	const struct nlattr *acts_if_greater, *acts_if_lesser_eq;
2953	struct nlattr *a[OVS_CHECK_PKT_LEN_ATTR_MAX + 1];
2954	struct check_pkt_len_arg arg;
2955	int nested_acts_start;
2956	int start, err;
2957
2958	err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
2959					  nla_data(attr), nla_len(attr),
2960					  cpl_policy, NULL);
2961	if (err)
2962		return err;
2963
2964	if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
2965	    !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
2966		return -EINVAL;
2967
2968	acts_if_lesser_eq = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL];
2969	acts_if_greater = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER];
2970
2971	/* Both the nested action should be present. */
2972	if (!acts_if_greater || !acts_if_lesser_eq)
2973		return -EINVAL;
2974
2975	/* validation done, copy the nested actions. */
2976	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
2977					log);
2978	if (start < 0)
2979		return start;
2980
2981	arg.pkt_len = nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]);
2982	arg.exec_for_lesser_equal =
2983		last || !actions_may_change_flow(acts_if_lesser_eq);
2984	arg.exec_for_greater =
2985		last || !actions_may_change_flow(acts_if_greater);
2986
2987	err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
2988				 sizeof(arg), log);
2989	if (err)
2990		return err;
2991
2992	nested_acts_start = add_nested_action_start(sfa,
2993		OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL, log);
2994	if (nested_acts_start < 0)
2995		return nested_acts_start;
2996
2997	err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, key, sfa,
2998				     eth_type, vlan_tci, mpls_label_count, log);
2999
3000	if (err)
3001		return err;
3002
3003	add_nested_action_end(*sfa, nested_acts_start);
3004
3005	nested_acts_start = add_nested_action_start(sfa,
3006		OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER, log);
3007	if (nested_acts_start < 0)
3008		return nested_acts_start;
3009
3010	err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3011				     eth_type, vlan_tci, mpls_label_count, log);
3012
3013	if (err)
3014		return err;
3015
3016	add_nested_action_end(*sfa, nested_acts_start);
3017	add_nested_action_end(*sfa, start);
3018	return 0;
3019}
3020
3021static int copy_action(const struct nlattr *from,
3022		       struct sw_flow_actions **sfa, bool log)
3023{
3024	int totlen = NLA_ALIGN(from->nla_len);
3025	struct nlattr *to;
3026
3027	to = reserve_sfa_size(sfa, from->nla_len, log);
3028	if (IS_ERR(to))
3029		return PTR_ERR(to);
3030
3031	memcpy(to, from, totlen);
3032	return 0;
3033}
3034
3035static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3036				  const struct sw_flow_key *key,
3037				  struct sw_flow_actions **sfa,
3038				  __be16 eth_type, __be16 vlan_tci,
3039				  u32 mpls_label_count, bool log)
3040{
3041	u8 mac_proto = ovs_key_mac_proto(key);
3042	const struct nlattr *a;
3043	int rem, err;
3044
3045	nla_for_each_nested(a, attr, rem) {
3046		/* Expected argument lengths, (u32)-1 for variable length. */
3047		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
3048			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
3049			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
3050			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
3051			[OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
3052			[OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
3053			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
3054			[OVS_ACTION_ATTR_POP_VLAN] = 0,
3055			[OVS_ACTION_ATTR_SET] = (u32)-1,
3056			[OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
3057			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
3058			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
3059			[OVS_ACTION_ATTR_CT] = (u32)-1,
3060			[OVS_ACTION_ATTR_CT_CLEAR] = 0,
3061			[OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
3062			[OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
3063			[OVS_ACTION_ATTR_POP_ETH] = 0,
3064			[OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
3065			[OVS_ACTION_ATTR_POP_NSH] = 0,
3066			[OVS_ACTION_ATTR_METER] = sizeof(u32),
3067			[OVS_ACTION_ATTR_CLONE] = (u32)-1,
3068			[OVS_ACTION_ATTR_CHECK_PKT_LEN] = (u32)-1,
3069			[OVS_ACTION_ATTR_ADD_MPLS] = sizeof(struct ovs_action_add_mpls),
3070			[OVS_ACTION_ATTR_DEC_TTL] = (u32)-1,
3071		};
3072		const struct ovs_action_push_vlan *vlan;
3073		int type = nla_type(a);
3074		bool skip_copy;
3075
3076		if (type > OVS_ACTION_ATTR_MAX ||
3077		    (action_lens[type] != nla_len(a) &&
3078		     action_lens[type] != (u32)-1))
3079			return -EINVAL;
3080
3081		skip_copy = false;
3082		switch (type) {
3083		case OVS_ACTION_ATTR_UNSPEC:
3084			return -EINVAL;
3085
3086		case OVS_ACTION_ATTR_USERSPACE:
3087			err = validate_userspace(a);
3088			if (err)
3089				return err;
3090			break;
3091
3092		case OVS_ACTION_ATTR_OUTPUT:
3093			if (nla_get_u32(a) >= DP_MAX_PORTS)
3094				return -EINVAL;
3095			break;
3096
3097		case OVS_ACTION_ATTR_TRUNC: {
3098			const struct ovs_action_trunc *trunc = nla_data(a);
3099
3100			if (trunc->max_len < ETH_HLEN)
3101				return -EINVAL;
3102			break;
3103		}
3104
3105		case OVS_ACTION_ATTR_HASH: {
3106			const struct ovs_action_hash *act_hash = nla_data(a);
3107
3108			switch (act_hash->hash_alg) {
3109			case OVS_HASH_ALG_L4:
3110				break;
3111			default:
3112				return  -EINVAL;
3113			}
3114
3115			break;
3116		}
3117
3118		case OVS_ACTION_ATTR_POP_VLAN:
3119			if (mac_proto != MAC_PROTO_ETHERNET)
3120				return -EINVAL;
3121			vlan_tci = htons(0);
3122			break;
3123
3124		case OVS_ACTION_ATTR_PUSH_VLAN:
3125			if (mac_proto != MAC_PROTO_ETHERNET)
3126				return -EINVAL;
3127			vlan = nla_data(a);
3128			if (!eth_type_vlan(vlan->vlan_tpid))
3129				return -EINVAL;
3130			if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3131				return -EINVAL;
3132			vlan_tci = vlan->vlan_tci;
3133			break;
3134
3135		case OVS_ACTION_ATTR_RECIRC:
3136			break;
3137
3138		case OVS_ACTION_ATTR_ADD_MPLS: {
3139			const struct ovs_action_add_mpls *mpls = nla_data(a);
3140
3141			if (!eth_p_mpls(mpls->mpls_ethertype))
3142				return -EINVAL;
3143
3144			if (mpls->tun_flags & OVS_MPLS_L3_TUNNEL_FLAG_MASK) {
3145				if (vlan_tci & htons(VLAN_CFI_MASK) ||
3146				    (eth_type != htons(ETH_P_IP) &&
3147				     eth_type != htons(ETH_P_IPV6) &&
3148				     eth_type != htons(ETH_P_ARP) &&
3149				     eth_type != htons(ETH_P_RARP) &&
3150				     !eth_p_mpls(eth_type)))
3151					return -EINVAL;
3152				mpls_label_count++;
3153			} else {
3154				if (mac_proto == MAC_PROTO_ETHERNET) {
3155					mpls_label_count = 1;
3156					mac_proto = MAC_PROTO_NONE;
3157				} else {
3158					mpls_label_count++;
3159				}
3160			}
3161			eth_type = mpls->mpls_ethertype;
3162			break;
3163		}
3164
3165		case OVS_ACTION_ATTR_PUSH_MPLS: {
3166			const struct ovs_action_push_mpls *mpls = nla_data(a);
3167
3168			if (!eth_p_mpls(mpls->mpls_ethertype))
3169				return -EINVAL;
3170			/* Prohibit push MPLS other than to a white list
3171			 * for packets that have a known tag order.
3172			 */
3173			if (vlan_tci & htons(VLAN_CFI_MASK) ||
3174			    (eth_type != htons(ETH_P_IP) &&
3175			     eth_type != htons(ETH_P_IPV6) &&
3176			     eth_type != htons(ETH_P_ARP) &&
3177			     eth_type != htons(ETH_P_RARP) &&
3178			     !eth_p_mpls(eth_type)))
3179				return -EINVAL;
3180			eth_type = mpls->mpls_ethertype;
3181			mpls_label_count++;
3182			break;
3183		}
3184
3185		case OVS_ACTION_ATTR_POP_MPLS: {
3186			__be16  proto;
3187			if (vlan_tci & htons(VLAN_CFI_MASK) ||
3188			    !eth_p_mpls(eth_type))
3189				return -EINVAL;
3190
3191			/* Disallow subsequent L2.5+ set actions and mpls_pop
3192			 * actions once the last MPLS label in the packet is
3193			 * is popped as there is no check here to ensure that
3194			 * the new eth type is valid and thus set actions could
3195			 * write off the end of the packet or otherwise corrupt
3196			 * it.
3197			 *
3198			 * Support for these actions is planned using packet
3199			 * recirculation.
3200			 */
3201			proto = nla_get_be16(a);
3202
3203			if (proto == htons(ETH_P_TEB) &&
3204			    mac_proto != MAC_PROTO_NONE)
3205				return -EINVAL;
3206
3207			mpls_label_count--;
3208
3209			if (!eth_p_mpls(proto) || !mpls_label_count)
3210				eth_type = htons(0);
3211			else
3212				eth_type =  proto;
3213
3214			break;
3215		}
3216
3217		case OVS_ACTION_ATTR_SET:
3218			err = validate_set(a, key, sfa,
3219					   &skip_copy, mac_proto, eth_type,
3220					   false, log);
3221			if (err)
3222				return err;
3223			break;
3224
3225		case OVS_ACTION_ATTR_SET_MASKED:
3226			err = validate_set(a, key, sfa,
3227					   &skip_copy, mac_proto, eth_type,
3228					   true, log);
3229			if (err)
3230				return err;
3231			break;
3232
3233		case OVS_ACTION_ATTR_SAMPLE: {
3234			bool last = nla_is_last(a, rem);
3235
3236			err = validate_and_copy_sample(net, a, key, sfa,
3237						       eth_type, vlan_tci,
3238						       mpls_label_count,
3239						       log, last);
3240			if (err)
3241				return err;
3242			skip_copy = true;
3243			break;
3244		}
3245
3246		case OVS_ACTION_ATTR_CT:
3247			err = ovs_ct_copy_action(net, a, key, sfa, log);
3248			if (err)
3249				return err;
3250			skip_copy = true;
3251			break;
3252
3253		case OVS_ACTION_ATTR_CT_CLEAR:
3254			break;
3255
3256		case OVS_ACTION_ATTR_PUSH_ETH:
3257			/* Disallow pushing an Ethernet header if one
3258			 * is already present */
3259			if (mac_proto != MAC_PROTO_NONE)
3260				return -EINVAL;
3261			mac_proto = MAC_PROTO_ETHERNET;
3262			break;
3263
3264		case OVS_ACTION_ATTR_POP_ETH:
3265			if (mac_proto != MAC_PROTO_ETHERNET)
3266				return -EINVAL;
3267			if (vlan_tci & htons(VLAN_CFI_MASK))
3268				return -EINVAL;
3269			mac_proto = MAC_PROTO_NONE;
3270			break;
3271
3272		case OVS_ACTION_ATTR_PUSH_NSH:
3273			if (mac_proto != MAC_PROTO_ETHERNET) {
3274				u8 next_proto;
3275
3276				next_proto = tun_p_from_eth_p(eth_type);
3277				if (!next_proto)
3278					return -EINVAL;
3279			}
3280			mac_proto = MAC_PROTO_NONE;
3281			if (!validate_nsh(nla_data(a), false, true, true))
3282				return -EINVAL;
3283			break;
3284
3285		case OVS_ACTION_ATTR_POP_NSH: {
3286			__be16 inner_proto;
3287
3288			if (eth_type != htons(ETH_P_NSH))
3289				return -EINVAL;
3290			inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3291			if (!inner_proto)
3292				return -EINVAL;
3293			if (key->nsh.base.np == TUN_P_ETHERNET)
3294				mac_proto = MAC_PROTO_ETHERNET;
3295			else
3296				mac_proto = MAC_PROTO_NONE;
3297			break;
3298		}
3299
3300		case OVS_ACTION_ATTR_METER:
3301			/* Non-existent meters are simply ignored.  */
3302			break;
3303
3304		case OVS_ACTION_ATTR_CLONE: {
3305			bool last = nla_is_last(a, rem);
3306
3307			err = validate_and_copy_clone(net, a, key, sfa,
3308						      eth_type, vlan_tci,
3309						      mpls_label_count,
3310						      log, last);
3311			if (err)
3312				return err;
3313			skip_copy = true;
3314			break;
3315		}
3316
3317		case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3318			bool last = nla_is_last(a, rem);
3319
3320			err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3321							      eth_type,
3322							      vlan_tci,
3323							      mpls_label_count,
3324							      log, last);
3325			if (err)
3326				return err;
3327			skip_copy = true;
3328			break;
3329		}
3330
3331		case OVS_ACTION_ATTR_DEC_TTL:
3332			err = validate_and_copy_dec_ttl(net, a, key, sfa,
3333							eth_type, vlan_tci,
3334							mpls_label_count, log);
3335			if (err)
3336				return err;
3337			skip_copy = true;
3338			break;
3339
3340		default:
3341			OVS_NLERR(log, "Unknown Action type %d", type);
3342			return -EINVAL;
3343		}
3344		if (!skip_copy) {
3345			err = copy_action(a, sfa, log);
3346			if (err)
3347				return err;
3348		}
3349	}
3350
3351	if (rem > 0)
3352		return -EINVAL;
3353
3354	return 0;
3355}
3356
3357/* 'key' must be the masked key. */
3358int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3359			 const struct sw_flow_key *key,
3360			 struct sw_flow_actions **sfa, bool log)
3361{
3362	int err;
3363	u32 mpls_label_count = 0;
3364
3365	*sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3366	if (IS_ERR(*sfa))
3367		return PTR_ERR(*sfa);
3368
3369	if (eth_p_mpls(key->eth.type))
3370		mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3371
3372	(*sfa)->orig_len = nla_len(attr);
3373	err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3374				     key->eth.vlan.tci, mpls_label_count, log);
3375	if (err)
3376		ovs_nla_free_flow_actions(*sfa);
3377
3378	return err;
3379}
3380
3381static int sample_action_to_attr(const struct nlattr *attr,
3382				 struct sk_buff *skb)
3383{
3384	struct nlattr *start, *ac_start = NULL, *sample_arg;
3385	int err = 0, rem = nla_len(attr);
3386	const struct sample_arg *arg;
3387	struct nlattr *actions;
3388
3389	start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3390	if (!start)
3391		return -EMSGSIZE;
3392
3393	sample_arg = nla_data(attr);
3394	arg = nla_data(sample_arg);
3395	actions = nla_next(sample_arg, &rem);
3396
3397	if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3398		err = -EMSGSIZE;
3399		goto out;
3400	}
3401
3402	ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3403	if (!ac_start) {
3404		err = -EMSGSIZE;
3405		goto out;
3406	}
3407
3408	err = ovs_nla_put_actions(actions, rem, skb);
3409
3410out:
3411	if (err) {
3412		nla_nest_cancel(skb, ac_start);
3413		nla_nest_cancel(skb, start);
3414	} else {
3415		nla_nest_end(skb, ac_start);
3416		nla_nest_end(skb, start);
3417	}
3418
3419	return err;
3420}
3421
3422static int clone_action_to_attr(const struct nlattr *attr,
3423				struct sk_buff *skb)
3424{
3425	struct nlattr *start;
3426	int err = 0, rem = nla_len(attr);
3427
3428	start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3429	if (!start)
3430		return -EMSGSIZE;
3431
3432	err = ovs_nla_put_actions(nla_data(attr), rem, skb);
3433
3434	if (err)
3435		nla_nest_cancel(skb, start);
3436	else
3437		nla_nest_end(skb, start);
3438
3439	return err;
3440}
3441
3442static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3443					struct sk_buff *skb)
3444{
3445	struct nlattr *start, *ac_start = NULL;
3446	const struct check_pkt_len_arg *arg;
3447	const struct nlattr *a, *cpl_arg;
3448	int err = 0, rem = nla_len(attr);
3449
3450	start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3451	if (!start)
3452		return -EMSGSIZE;
3453
3454	/* The first nested attribute in 'attr' is always
3455	 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3456	 */
3457	cpl_arg = nla_data(attr);
3458	arg = nla_data(cpl_arg);
3459
3460	if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3461		err = -EMSGSIZE;
3462		goto out;
3463	}
3464
3465	/* Second nested attribute in 'attr' is always
3466	 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3467	 */
3468	a = nla_next(cpl_arg, &rem);
3469	ac_start =  nla_nest_start_noflag(skb,
3470					  OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL);
3471	if (!ac_start) {
3472		err = -EMSGSIZE;
3473		goto out;
3474	}
3475
3476	err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3477	if (err) {
3478		nla_nest_cancel(skb, ac_start);
3479		goto out;
3480	} else {
3481		nla_nest_end(skb, ac_start);
3482	}
3483
3484	/* Third nested attribute in 'attr' is always
3485	 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3486	 */
3487	a = nla_next(a, &rem);
3488	ac_start =  nla_nest_start_noflag(skb,
3489					  OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3490	if (!ac_start) {
3491		err = -EMSGSIZE;
3492		goto out;
3493	}
3494
3495	err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3496	if (err) {
3497		nla_nest_cancel(skb, ac_start);
3498		goto out;
3499	} else {
3500		nla_nest_end(skb, ac_start);
3501	}
3502
3503	nla_nest_end(skb, start);
3504	return 0;
3505
3506out:
3507	nla_nest_cancel(skb, start);
3508	return err;
3509}
3510
3511static int dec_ttl_action_to_attr(const struct nlattr *attr,
3512				  struct sk_buff *skb)
3513{
3514	struct nlattr *start, *action_start;
3515	const struct nlattr *a;
3516	int err = 0, rem;
3517
3518	start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3519	if (!start)
3520		return -EMSGSIZE;
3521
3522	nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
3523		switch (nla_type(a)) {
3524		case OVS_DEC_TTL_ATTR_ACTION:
3525
3526			action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3527			if (!action_start) {
3528				err = -EMSGSIZE;
3529				goto out;
3530			}
3531
3532			err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3533			if (err)
3534				goto out;
3535
3536			nla_nest_end(skb, action_start);
3537			break;
3538
3539		default:
3540			/* Ignore all other option to be future compatible */
3541			break;
3542		}
3543	}
3544
3545	nla_nest_end(skb, start);
3546	return 0;
3547
3548out:
3549	nla_nest_cancel(skb, start);
3550	return err;
3551}
3552
3553static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3554{
3555	const struct nlattr *ovs_key = nla_data(a);
3556	int key_type = nla_type(ovs_key);
3557	struct nlattr *start;
3558	int err;
3559
3560	switch (key_type) {
3561	case OVS_KEY_ATTR_TUNNEL_INFO: {
3562		struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3563		struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3564
3565		start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3566		if (!start)
3567			return -EMSGSIZE;
3568
3569		err =  ip_tun_to_nlattr(skb, &tun_info->key,
3570					ip_tunnel_info_opts(tun_info),
3571					tun_info->options_len,
3572					ip_tunnel_info_af(tun_info), tun_info->mode);
3573		if (err)
3574			return err;
3575		nla_nest_end(skb, start);
3576		break;
3577	}
3578	default:
3579		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3580			return -EMSGSIZE;
3581		break;
3582	}
3583
3584	return 0;
3585}
3586
3587static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3588						struct sk_buff *skb)
3589{
3590	const struct nlattr *ovs_key = nla_data(a);
3591	struct nlattr *nla;
3592	size_t key_len = nla_len(ovs_key) / 2;
3593
3594	/* Revert the conversion we did from a non-masked set action to
3595	 * masked set action.
3596	 */
3597	nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3598	if (!nla)
3599		return -EMSGSIZE;
3600
3601	if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3602		return -EMSGSIZE;
3603
3604	nla_nest_end(skb, nla);
3605	return 0;
3606}
3607
3608int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3609{
3610	const struct nlattr *a;
3611	int rem, err;
3612
3613	nla_for_each_attr(a, attr, len, rem) {
3614		int type = nla_type(a);
3615
3616		switch (type) {
3617		case OVS_ACTION_ATTR_SET:
3618			err = set_action_to_attr(a, skb);
3619			if (err)
3620				return err;
3621			break;
3622
3623		case OVS_ACTION_ATTR_SET_TO_MASKED:
3624			err = masked_set_action_to_set_action_attr(a, skb);
3625			if (err)
3626				return err;
3627			break;
3628
3629		case OVS_ACTION_ATTR_SAMPLE:
3630			err = sample_action_to_attr(a, skb);
3631			if (err)
3632				return err;
3633			break;
3634
3635		case OVS_ACTION_ATTR_CT:
3636			err = ovs_ct_action_to_attr(nla_data(a), skb);
3637			if (err)
3638				return err;
3639			break;
3640
3641		case OVS_ACTION_ATTR_CLONE:
3642			err = clone_action_to_attr(a, skb);
3643			if (err)
3644				return err;
3645			break;
3646
3647		case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3648			err = check_pkt_len_action_to_attr(a, skb);
3649			if (err)
3650				return err;
3651			break;
3652
3653		case OVS_ACTION_ATTR_DEC_TTL:
3654			err = dec_ttl_action_to_attr(a, skb);
3655			if (err)
3656				return err;
3657			break;
3658
3659		default:
3660			if (nla_put(skb, type, nla_len(a), nla_data(a)))
3661				return -EMSGSIZE;
3662			break;
3663		}
3664	}
3665
3666	return 0;
3667}
v4.17
 
   1/*
   2 * Copyright (c) 2007-2017 Nicira, Inc.
   3 *
   4 * This program is free software; you can redistribute it and/or
   5 * modify it under the terms of version 2 of the GNU General Public
   6 * License as published by the Free Software Foundation.
   7 *
   8 * This program is distributed in the hope that it will be useful, but
   9 * WITHOUT ANY WARRANTY; without even the implied warranty of
  10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11 * General Public License for more details.
  12 *
  13 * You should have received a copy of the GNU General Public License
  14 * along with this program; if not, write to the Free Software
  15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16 * 02110-1301, USA
  17 */
  18
  19#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  20
  21#include "flow.h"
  22#include "datapath.h"
  23#include <linux/uaccess.h>
  24#include <linux/netdevice.h>
  25#include <linux/etherdevice.h>
  26#include <linux/if_ether.h>
  27#include <linux/if_vlan.h>
  28#include <net/llc_pdu.h>
  29#include <linux/kernel.h>
  30#include <linux/jhash.h>
  31#include <linux/jiffies.h>
  32#include <linux/llc.h>
  33#include <linux/module.h>
  34#include <linux/in.h>
  35#include <linux/rcupdate.h>
  36#include <linux/if_arp.h>
  37#include <linux/ip.h>
  38#include <linux/ipv6.h>
  39#include <linux/sctp.h>
  40#include <linux/tcp.h>
  41#include <linux/udp.h>
  42#include <linux/icmp.h>
  43#include <linux/icmpv6.h>
  44#include <linux/rculist.h>
  45#include <net/geneve.h>
  46#include <net/ip.h>
  47#include <net/ipv6.h>
  48#include <net/ndisc.h>
  49#include <net/mpls.h>
  50#include <net/vxlan.h>
  51#include <net/tun_proto.h>
  52#include <net/erspan.h>
  53
  54#include "flow_netlink.h"
  55
  56struct ovs_len_tbl {
  57	int len;
  58	const struct ovs_len_tbl *next;
  59};
  60
  61#define OVS_ATTR_NESTED -1
  62#define OVS_ATTR_VARIABLE -2
  63
  64static bool actions_may_change_flow(const struct nlattr *actions)
  65{
  66	struct nlattr *nla;
  67	int rem;
  68
  69	nla_for_each_nested(nla, actions, rem) {
  70		u16 action = nla_type(nla);
  71
  72		switch (action) {
  73		case OVS_ACTION_ATTR_OUTPUT:
  74		case OVS_ACTION_ATTR_RECIRC:
  75		case OVS_ACTION_ATTR_TRUNC:
  76		case OVS_ACTION_ATTR_USERSPACE:
  77			break;
  78
  79		case OVS_ACTION_ATTR_CT:
  80		case OVS_ACTION_ATTR_CT_CLEAR:
  81		case OVS_ACTION_ATTR_HASH:
  82		case OVS_ACTION_ATTR_POP_ETH:
  83		case OVS_ACTION_ATTR_POP_MPLS:
  84		case OVS_ACTION_ATTR_POP_NSH:
  85		case OVS_ACTION_ATTR_POP_VLAN:
  86		case OVS_ACTION_ATTR_PUSH_ETH:
  87		case OVS_ACTION_ATTR_PUSH_MPLS:
  88		case OVS_ACTION_ATTR_PUSH_NSH:
  89		case OVS_ACTION_ATTR_PUSH_VLAN:
  90		case OVS_ACTION_ATTR_SAMPLE:
  91		case OVS_ACTION_ATTR_SET:
  92		case OVS_ACTION_ATTR_SET_MASKED:
  93		case OVS_ACTION_ATTR_METER:
 
 
 
  94		default:
  95			return true;
  96		}
  97	}
  98	return false;
  99}
 100
 101static void update_range(struct sw_flow_match *match,
 102			 size_t offset, size_t size, bool is_mask)
 103{
 104	struct sw_flow_key_range *range;
 105	size_t start = rounddown(offset, sizeof(long));
 106	size_t end = roundup(offset + size, sizeof(long));
 107
 108	if (!is_mask)
 109		range = &match->range;
 110	else
 111		range = &match->mask->range;
 112
 113	if (range->start == range->end) {
 114		range->start = start;
 115		range->end = end;
 116		return;
 117	}
 118
 119	if (range->start > start)
 120		range->start = start;
 121
 122	if (range->end < end)
 123		range->end = end;
 124}
 125
 126#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
 127	do { \
 128		update_range(match, offsetof(struct sw_flow_key, field),    \
 129			     sizeof((match)->key->field), is_mask);	    \
 130		if (is_mask)						    \
 131			(match)->mask->key.field = value;		    \
 132		else							    \
 133			(match)->key->field = value;		            \
 134	} while (0)
 135
 136#define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask)	    \
 137	do {								    \
 138		update_range(match, offset, len, is_mask);		    \
 139		if (is_mask)						    \
 140			memcpy((u8 *)&(match)->mask->key + offset, value_p, \
 141			       len);					   \
 142		else							    \
 143			memcpy((u8 *)(match)->key + offset, value_p, len);  \
 144	} while (0)
 145
 146#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask)		      \
 147	SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
 148				  value_p, len, is_mask)
 149
 150#define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask)		    \
 151	do {								    \
 152		update_range(match, offsetof(struct sw_flow_key, field),    \
 153			     sizeof((match)->key->field), is_mask);	    \
 154		if (is_mask)						    \
 155			memset((u8 *)&(match)->mask->key.field, value,      \
 156			       sizeof((match)->mask->key.field));	    \
 157		else							    \
 158			memset((u8 *)&(match)->key->field, value,           \
 159			       sizeof((match)->key->field));                \
 160	} while (0)
 161
 162static bool match_validate(const struct sw_flow_match *match,
 163			   u64 key_attrs, u64 mask_attrs, bool log)
 164{
 165	u64 key_expected = 0;
 166	u64 mask_allowed = key_attrs;  /* At most allow all key attributes */
 167
 168	/* The following mask attributes allowed only if they
 169	 * pass the validation tests. */
 170	mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
 171			| (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
 172			| (1 << OVS_KEY_ATTR_IPV6)
 173			| (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
 174			| (1 << OVS_KEY_ATTR_TCP)
 175			| (1 << OVS_KEY_ATTR_TCP_FLAGS)
 176			| (1 << OVS_KEY_ATTR_UDP)
 177			| (1 << OVS_KEY_ATTR_SCTP)
 178			| (1 << OVS_KEY_ATTR_ICMP)
 179			| (1 << OVS_KEY_ATTR_ICMPV6)
 180			| (1 << OVS_KEY_ATTR_ARP)
 181			| (1 << OVS_KEY_ATTR_ND)
 182			| (1 << OVS_KEY_ATTR_MPLS)
 183			| (1 << OVS_KEY_ATTR_NSH));
 184
 185	/* Always allowed mask fields. */
 186	mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
 187		       | (1 << OVS_KEY_ATTR_IN_PORT)
 188		       | (1 << OVS_KEY_ATTR_ETHERTYPE));
 189
 190	/* Check key attributes. */
 191	if (match->key->eth.type == htons(ETH_P_ARP)
 192			|| match->key->eth.type == htons(ETH_P_RARP)) {
 193		key_expected |= 1 << OVS_KEY_ATTR_ARP;
 194		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
 195			mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
 196	}
 197
 198	if (eth_p_mpls(match->key->eth.type)) {
 199		key_expected |= 1 << OVS_KEY_ATTR_MPLS;
 200		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
 201			mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
 202	}
 203
 204	if (match->key->eth.type == htons(ETH_P_IP)) {
 205		key_expected |= 1 << OVS_KEY_ATTR_IPV4;
 206		if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
 207			mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
 208			mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
 209		}
 210
 211		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
 212			if (match->key->ip.proto == IPPROTO_UDP) {
 213				key_expected |= 1 << OVS_KEY_ATTR_UDP;
 214				if (match->mask && (match->mask->key.ip.proto == 0xff))
 215					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
 216			}
 217
 218			if (match->key->ip.proto == IPPROTO_SCTP) {
 219				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
 220				if (match->mask && (match->mask->key.ip.proto == 0xff))
 221					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
 222			}
 223
 224			if (match->key->ip.proto == IPPROTO_TCP) {
 225				key_expected |= 1 << OVS_KEY_ATTR_TCP;
 226				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 227				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
 228					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
 229					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 230				}
 231			}
 232
 233			if (match->key->ip.proto == IPPROTO_ICMP) {
 234				key_expected |= 1 << OVS_KEY_ATTR_ICMP;
 235				if (match->mask && (match->mask->key.ip.proto == 0xff))
 236					mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
 237			}
 238		}
 239	}
 240
 241	if (match->key->eth.type == htons(ETH_P_IPV6)) {
 242		key_expected |= 1 << OVS_KEY_ATTR_IPV6;
 243		if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
 244			mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
 245			mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
 246		}
 247
 248		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
 249			if (match->key->ip.proto == IPPROTO_UDP) {
 250				key_expected |= 1 << OVS_KEY_ATTR_UDP;
 251				if (match->mask && (match->mask->key.ip.proto == 0xff))
 252					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
 253			}
 254
 255			if (match->key->ip.proto == IPPROTO_SCTP) {
 256				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
 257				if (match->mask && (match->mask->key.ip.proto == 0xff))
 258					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
 259			}
 260
 261			if (match->key->ip.proto == IPPROTO_TCP) {
 262				key_expected |= 1 << OVS_KEY_ATTR_TCP;
 263				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 264				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
 265					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
 266					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
 267				}
 268			}
 269
 270			if (match->key->ip.proto == IPPROTO_ICMPV6) {
 271				key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
 272				if (match->mask && (match->mask->key.ip.proto == 0xff))
 273					mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
 274
 275				if (match->key->tp.src ==
 276						htons(NDISC_NEIGHBOUR_SOLICITATION) ||
 277				    match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
 278					key_expected |= 1 << OVS_KEY_ATTR_ND;
 279					/* Original direction conntrack tuple
 280					 * uses the same space as the ND fields
 281					 * in the key, so both are not allowed
 282					 * at the same time.
 283					 */
 284					mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
 285					if (match->mask && (match->mask->key.tp.src == htons(0xff)))
 286						mask_allowed |= 1 << OVS_KEY_ATTR_ND;
 287				}
 288			}
 289		}
 290	}
 291
 292	if (match->key->eth.type == htons(ETH_P_NSH)) {
 293		key_expected |= 1 << OVS_KEY_ATTR_NSH;
 294		if (match->mask &&
 295		    match->mask->key.eth.type == htons(0xffff)) {
 296			mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
 297		}
 298	}
 299
 300	if ((key_attrs & key_expected) != key_expected) {
 301		/* Key attributes check failed. */
 302		OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
 303			  (unsigned long long)key_attrs,
 304			  (unsigned long long)key_expected);
 305		return false;
 306	}
 307
 308	if ((mask_attrs & mask_allowed) != mask_attrs) {
 309		/* Mask attributes check failed. */
 310		OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
 311			  (unsigned long long)mask_attrs,
 312			  (unsigned long long)mask_allowed);
 313		return false;
 314	}
 315
 316	return true;
 317}
 318
 319size_t ovs_tun_key_attr_size(void)
 320{
 321	/* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
 322	 * updating this function.
 323	 */
 324	return    nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
 325		+ nla_total_size(16)   /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
 326		+ nla_total_size(16)   /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
 327		+ nla_total_size(1)    /* OVS_TUNNEL_KEY_ATTR_TOS */
 328		+ nla_total_size(1)    /* OVS_TUNNEL_KEY_ATTR_TTL */
 329		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
 330		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_CSUM */
 331		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_OAM */
 332		+ nla_total_size(256)  /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
 333		/* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
 334		 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
 335		 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
 336		 */
 337		+ nla_total_size(2)    /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
 338		+ nla_total_size(2);   /* OVS_TUNNEL_KEY_ATTR_TP_DST */
 339}
 340
 341static size_t ovs_nsh_key_attr_size(void)
 342{
 343	/* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
 344	 * updating this function.
 345	 */
 346	return  nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
 347		/* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
 348		 * mutually exclusive, so the bigger one can cover
 349		 * the small one.
 350		 */
 351		+ nla_total_size(NSH_CTX_HDRS_MAX_LEN);
 352}
 353
 354size_t ovs_key_attr_size(void)
 355{
 356	/* Whenever adding new OVS_KEY_ FIELDS, we should consider
 357	 * updating this function.
 358	 */
 359	BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 29);
 360
 361	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
 362		+ nla_total_size(0)   /* OVS_KEY_ATTR_TUNNEL */
 363		  + ovs_tun_key_attr_size()
 364		+ nla_total_size(4)   /* OVS_KEY_ATTR_IN_PORT */
 365		+ nla_total_size(4)   /* OVS_KEY_ATTR_SKB_MARK */
 366		+ nla_total_size(4)   /* OVS_KEY_ATTR_DP_HASH */
 367		+ nla_total_size(4)   /* OVS_KEY_ATTR_RECIRC_ID */
 368		+ nla_total_size(4)   /* OVS_KEY_ATTR_CT_STATE */
 369		+ nla_total_size(2)   /* OVS_KEY_ATTR_CT_ZONE */
 370		+ nla_total_size(4)   /* OVS_KEY_ATTR_CT_MARK */
 371		+ nla_total_size(16)  /* OVS_KEY_ATTR_CT_LABELS */
 372		+ nla_total_size(40)  /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
 373		+ nla_total_size(0)   /* OVS_KEY_ATTR_NSH */
 374		  + ovs_nsh_key_attr_size()
 375		+ nla_total_size(12)  /* OVS_KEY_ATTR_ETHERNET */
 376		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
 377		+ nla_total_size(4)   /* OVS_KEY_ATTR_VLAN */
 378		+ nla_total_size(0)   /* OVS_KEY_ATTR_ENCAP */
 379		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
 380		+ nla_total_size(40)  /* OVS_KEY_ATTR_IPV6 */
 381		+ nla_total_size(2)   /* OVS_KEY_ATTR_ICMPV6 */
 382		+ nla_total_size(28); /* OVS_KEY_ATTR_ND */
 383}
 384
 385static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
 386	[OVS_VXLAN_EXT_GBP]	    = { .len = sizeof(u32) },
 387};
 388
 389static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
 390	[OVS_TUNNEL_KEY_ATTR_ID]	    = { .len = sizeof(u64) },
 391	[OVS_TUNNEL_KEY_ATTR_IPV4_SRC]	    = { .len = sizeof(u32) },
 392	[OVS_TUNNEL_KEY_ATTR_IPV4_DST]	    = { .len = sizeof(u32) },
 393	[OVS_TUNNEL_KEY_ATTR_TOS]	    = { .len = 1 },
 394	[OVS_TUNNEL_KEY_ATTR_TTL]	    = { .len = 1 },
 395	[OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
 396	[OVS_TUNNEL_KEY_ATTR_CSUM]	    = { .len = 0 },
 397	[OVS_TUNNEL_KEY_ATTR_TP_SRC]	    = { .len = sizeof(u16) },
 398	[OVS_TUNNEL_KEY_ATTR_TP_DST]	    = { .len = sizeof(u16) },
 399	[OVS_TUNNEL_KEY_ATTR_OAM]	    = { .len = 0 },
 400	[OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS]   = { .len = OVS_ATTR_VARIABLE },
 401	[OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS]    = { .len = OVS_ATTR_NESTED,
 402						.next = ovs_vxlan_ext_key_lens },
 403	[OVS_TUNNEL_KEY_ATTR_IPV6_SRC]      = { .len = sizeof(struct in6_addr) },
 404	[OVS_TUNNEL_KEY_ATTR_IPV6_DST]      = { .len = sizeof(struct in6_addr) },
 405	[OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS]   = { .len = OVS_ATTR_VARIABLE },
 
 406};
 407
 408static const struct ovs_len_tbl
 409ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
 410	[OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
 411	[OVS_NSH_KEY_ATTR_MD1]  = { .len = sizeof(struct ovs_nsh_key_md1) },
 412	[OVS_NSH_KEY_ATTR_MD2]  = { .len = OVS_ATTR_VARIABLE },
 413};
 414
 415/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
 416static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
 417	[OVS_KEY_ATTR_ENCAP]	 = { .len = OVS_ATTR_NESTED },
 418	[OVS_KEY_ATTR_PRIORITY]	 = { .len = sizeof(u32) },
 419	[OVS_KEY_ATTR_IN_PORT]	 = { .len = sizeof(u32) },
 420	[OVS_KEY_ATTR_SKB_MARK]	 = { .len = sizeof(u32) },
 421	[OVS_KEY_ATTR_ETHERNET]	 = { .len = sizeof(struct ovs_key_ethernet) },
 422	[OVS_KEY_ATTR_VLAN]	 = { .len = sizeof(__be16) },
 423	[OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
 424	[OVS_KEY_ATTR_IPV4]	 = { .len = sizeof(struct ovs_key_ipv4) },
 425	[OVS_KEY_ATTR_IPV6]	 = { .len = sizeof(struct ovs_key_ipv6) },
 426	[OVS_KEY_ATTR_TCP]	 = { .len = sizeof(struct ovs_key_tcp) },
 427	[OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
 428	[OVS_KEY_ATTR_UDP]	 = { .len = sizeof(struct ovs_key_udp) },
 429	[OVS_KEY_ATTR_SCTP]	 = { .len = sizeof(struct ovs_key_sctp) },
 430	[OVS_KEY_ATTR_ICMP]	 = { .len = sizeof(struct ovs_key_icmp) },
 431	[OVS_KEY_ATTR_ICMPV6]	 = { .len = sizeof(struct ovs_key_icmpv6) },
 432	[OVS_KEY_ATTR_ARP]	 = { .len = sizeof(struct ovs_key_arp) },
 433	[OVS_KEY_ATTR_ND]	 = { .len = sizeof(struct ovs_key_nd) },
 434	[OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
 435	[OVS_KEY_ATTR_DP_HASH]	 = { .len = sizeof(u32) },
 436	[OVS_KEY_ATTR_TUNNEL]	 = { .len = OVS_ATTR_NESTED,
 437				     .next = ovs_tunnel_key_lens, },
 438	[OVS_KEY_ATTR_MPLS]	 = { .len = sizeof(struct ovs_key_mpls) },
 439	[OVS_KEY_ATTR_CT_STATE]	 = { .len = sizeof(u32) },
 440	[OVS_KEY_ATTR_CT_ZONE]	 = { .len = sizeof(u16) },
 441	[OVS_KEY_ATTR_CT_MARK]	 = { .len = sizeof(u32) },
 442	[OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
 443	[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
 444		.len = sizeof(struct ovs_key_ct_tuple_ipv4) },
 445	[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
 446		.len = sizeof(struct ovs_key_ct_tuple_ipv6) },
 447	[OVS_KEY_ATTR_NSH]       = { .len = OVS_ATTR_NESTED,
 448				     .next = ovs_nsh_key_attr_lens, },
 449};
 450
 451static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
 452{
 453	return expected_len == attr_len ||
 454	       expected_len == OVS_ATTR_NESTED ||
 455	       expected_len == OVS_ATTR_VARIABLE;
 456}
 457
 458static bool is_all_zero(const u8 *fp, size_t size)
 459{
 460	int i;
 461
 462	if (!fp)
 463		return false;
 464
 465	for (i = 0; i < size; i++)
 466		if (fp[i])
 467			return false;
 468
 469	return true;
 470}
 471
 472static int __parse_flow_nlattrs(const struct nlattr *attr,
 473				const struct nlattr *a[],
 474				u64 *attrsp, bool log, bool nz)
 475{
 476	const struct nlattr *nla;
 477	u64 attrs;
 478	int rem;
 479
 480	attrs = *attrsp;
 481	nla_for_each_nested(nla, attr, rem) {
 482		u16 type = nla_type(nla);
 483		int expected_len;
 484
 485		if (type > OVS_KEY_ATTR_MAX) {
 486			OVS_NLERR(log, "Key type %d is out of range max %d",
 487				  type, OVS_KEY_ATTR_MAX);
 488			return -EINVAL;
 489		}
 490
 491		if (attrs & (1 << type)) {
 492			OVS_NLERR(log, "Duplicate key (type %d).", type);
 493			return -EINVAL;
 494		}
 495
 496		expected_len = ovs_key_lens[type].len;
 497		if (!check_attr_len(nla_len(nla), expected_len)) {
 498			OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
 499				  type, nla_len(nla), expected_len);
 500			return -EINVAL;
 501		}
 502
 503		if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
 504			attrs |= 1 << type;
 505			a[type] = nla;
 506		}
 507	}
 508	if (rem) {
 509		OVS_NLERR(log, "Message has %d unknown bytes.", rem);
 510		return -EINVAL;
 511	}
 512
 513	*attrsp = attrs;
 514	return 0;
 515}
 516
 517static int parse_flow_mask_nlattrs(const struct nlattr *attr,
 518				   const struct nlattr *a[], u64 *attrsp,
 519				   bool log)
 520{
 521	return __parse_flow_nlattrs(attr, a, attrsp, log, true);
 522}
 523
 524int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
 525		       u64 *attrsp, bool log)
 526{
 527	return __parse_flow_nlattrs(attr, a, attrsp, log, false);
 528}
 529
 530static int genev_tun_opt_from_nlattr(const struct nlattr *a,
 531				     struct sw_flow_match *match, bool is_mask,
 532				     bool log)
 533{
 534	unsigned long opt_key_offset;
 535
 536	if (nla_len(a) > sizeof(match->key->tun_opts)) {
 537		OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
 538			  nla_len(a), sizeof(match->key->tun_opts));
 539		return -EINVAL;
 540	}
 541
 542	if (nla_len(a) % 4 != 0) {
 543		OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
 544			  nla_len(a));
 545		return -EINVAL;
 546	}
 547
 548	/* We need to record the length of the options passed
 549	 * down, otherwise packets with the same format but
 550	 * additional options will be silently matched.
 551	 */
 552	if (!is_mask) {
 553		SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
 554				false);
 555	} else {
 556		/* This is somewhat unusual because it looks at
 557		 * both the key and mask while parsing the
 558		 * attributes (and by extension assumes the key
 559		 * is parsed first). Normally, we would verify
 560		 * that each is the correct length and that the
 561		 * attributes line up in the validate function.
 562		 * However, that is difficult because this is
 563		 * variable length and we won't have the
 564		 * information later.
 565		 */
 566		if (match->key->tun_opts_len != nla_len(a)) {
 567			OVS_NLERR(log, "Geneve option len %d != mask len %d",
 568				  match->key->tun_opts_len, nla_len(a));
 569			return -EINVAL;
 570		}
 571
 572		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
 573	}
 574
 575	opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
 576	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
 577				  nla_len(a), is_mask);
 578	return 0;
 579}
 580
 581static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
 582				     struct sw_flow_match *match, bool is_mask,
 583				     bool log)
 584{
 585	struct nlattr *a;
 586	int rem;
 587	unsigned long opt_key_offset;
 588	struct vxlan_metadata opts;
 589
 590	BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
 591
 592	memset(&opts, 0, sizeof(opts));
 593	nla_for_each_nested(a, attr, rem) {
 594		int type = nla_type(a);
 595
 596		if (type > OVS_VXLAN_EXT_MAX) {
 597			OVS_NLERR(log, "VXLAN extension %d out of range max %d",
 598				  type, OVS_VXLAN_EXT_MAX);
 599			return -EINVAL;
 600		}
 601
 602		if (!check_attr_len(nla_len(a),
 603				    ovs_vxlan_ext_key_lens[type].len)) {
 604			OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
 605				  type, nla_len(a),
 606				  ovs_vxlan_ext_key_lens[type].len);
 607			return -EINVAL;
 608		}
 609
 610		switch (type) {
 611		case OVS_VXLAN_EXT_GBP:
 612			opts.gbp = nla_get_u32(a);
 613			break;
 614		default:
 615			OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
 616				  type);
 617			return -EINVAL;
 618		}
 619	}
 620	if (rem) {
 621		OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
 622			  rem);
 623		return -EINVAL;
 624	}
 625
 626	if (!is_mask)
 627		SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
 628	else
 629		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
 630
 631	opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
 632	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
 633				  is_mask);
 634	return 0;
 635}
 636
 637static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
 638				      struct sw_flow_match *match, bool is_mask,
 639				      bool log)
 640{
 641	unsigned long opt_key_offset;
 642
 643	BUILD_BUG_ON(sizeof(struct erspan_metadata) >
 644		     sizeof(match->key->tun_opts));
 645
 646	if (nla_len(a) > sizeof(match->key->tun_opts)) {
 647		OVS_NLERR(log, "ERSPAN option length err (len %d, max %zu).",
 648			  nla_len(a), sizeof(match->key->tun_opts));
 649		return -EINVAL;
 650	}
 651
 652	if (!is_mask)
 653		SW_FLOW_KEY_PUT(match, tun_opts_len,
 654				sizeof(struct erspan_metadata), false);
 655	else
 656		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
 657
 658	opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
 659	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
 660				  nla_len(a), is_mask);
 661	return 0;
 662}
 663
 664static int ip_tun_from_nlattr(const struct nlattr *attr,
 665			      struct sw_flow_match *match, bool is_mask,
 666			      bool log)
 667{
 668	bool ttl = false, ipv4 = false, ipv6 = false;
 
 669	__be16 tun_flags = 0;
 670	int opts_type = 0;
 671	struct nlattr *a;
 672	int rem;
 673
 674	nla_for_each_nested(a, attr, rem) {
 675		int type = nla_type(a);
 676		int err;
 677
 678		if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
 679			OVS_NLERR(log, "Tunnel attr %d out of range max %d",
 680				  type, OVS_TUNNEL_KEY_ATTR_MAX);
 681			return -EINVAL;
 682		}
 683
 684		if (!check_attr_len(nla_len(a),
 685				    ovs_tunnel_key_lens[type].len)) {
 686			OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
 687				  type, nla_len(a), ovs_tunnel_key_lens[type].len);
 688			return -EINVAL;
 689		}
 690
 691		switch (type) {
 692		case OVS_TUNNEL_KEY_ATTR_ID:
 693			SW_FLOW_KEY_PUT(match, tun_key.tun_id,
 694					nla_get_be64(a), is_mask);
 695			tun_flags |= TUNNEL_KEY;
 696			break;
 697		case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
 698			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
 699					nla_get_in_addr(a), is_mask);
 700			ipv4 = true;
 701			break;
 702		case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
 703			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
 704					nla_get_in_addr(a), is_mask);
 705			ipv4 = true;
 706			break;
 707		case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
 708			SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
 709					nla_get_in6_addr(a), is_mask);
 710			ipv6 = true;
 711			break;
 712		case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
 713			SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
 714					nla_get_in6_addr(a), is_mask);
 715			ipv6 = true;
 716			break;
 717		case OVS_TUNNEL_KEY_ATTR_TOS:
 718			SW_FLOW_KEY_PUT(match, tun_key.tos,
 719					nla_get_u8(a), is_mask);
 720			break;
 721		case OVS_TUNNEL_KEY_ATTR_TTL:
 722			SW_FLOW_KEY_PUT(match, tun_key.ttl,
 723					nla_get_u8(a), is_mask);
 724			ttl = true;
 725			break;
 726		case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
 727			tun_flags |= TUNNEL_DONT_FRAGMENT;
 728			break;
 729		case OVS_TUNNEL_KEY_ATTR_CSUM:
 730			tun_flags |= TUNNEL_CSUM;
 731			break;
 732		case OVS_TUNNEL_KEY_ATTR_TP_SRC:
 733			SW_FLOW_KEY_PUT(match, tun_key.tp_src,
 734					nla_get_be16(a), is_mask);
 735			break;
 736		case OVS_TUNNEL_KEY_ATTR_TP_DST:
 737			SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
 738					nla_get_be16(a), is_mask);
 739			break;
 740		case OVS_TUNNEL_KEY_ATTR_OAM:
 741			tun_flags |= TUNNEL_OAM;
 742			break;
 743		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
 744			if (opts_type) {
 745				OVS_NLERR(log, "Multiple metadata blocks provided");
 746				return -EINVAL;
 747			}
 748
 749			err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
 750			if (err)
 751				return err;
 752
 753			tun_flags |= TUNNEL_GENEVE_OPT;
 754			opts_type = type;
 755			break;
 756		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
 757			if (opts_type) {
 758				OVS_NLERR(log, "Multiple metadata blocks provided");
 759				return -EINVAL;
 760			}
 761
 762			err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
 763			if (err)
 764				return err;
 765
 766			tun_flags |= TUNNEL_VXLAN_OPT;
 767			opts_type = type;
 768			break;
 769		case OVS_TUNNEL_KEY_ATTR_PAD:
 770			break;
 771		case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
 772			if (opts_type) {
 773				OVS_NLERR(log, "Multiple metadata blocks provided");
 774				return -EINVAL;
 775			}
 776
 777			err = erspan_tun_opt_from_nlattr(a, match, is_mask,
 778							 log);
 779			if (err)
 780				return err;
 781
 782			tun_flags |= TUNNEL_ERSPAN_OPT;
 783			opts_type = type;
 784			break;
 
 
 
 
 785		default:
 786			OVS_NLERR(log, "Unknown IP tunnel attribute %d",
 787				  type);
 788			return -EINVAL;
 789		}
 790	}
 791
 792	SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
 793	if (is_mask)
 794		SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
 795	else
 796		SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
 797				false);
 798
 799	if (rem > 0) {
 800		OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
 801			  rem);
 802		return -EINVAL;
 803	}
 804
 805	if (ipv4 && ipv6) {
 806		OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
 807		return -EINVAL;
 808	}
 809
 810	if (!is_mask) {
 811		if (!ipv4 && !ipv6) {
 812			OVS_NLERR(log, "IP tunnel dst address not specified");
 813			return -EINVAL;
 814		}
 815		if (ipv4 && !match->key->tun_key.u.ipv4.dst) {
 816			OVS_NLERR(log, "IPv4 tunnel dst address is zero");
 817			return -EINVAL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 818		}
 819		if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
 820			OVS_NLERR(log, "IPv6 tunnel dst address is zero");
 821			return -EINVAL;
 822		}
 823
 824		if (!ttl) {
 825			OVS_NLERR(log, "IP tunnel TTL not specified.");
 826			return -EINVAL;
 827		}
 828	}
 829
 830	return opts_type;
 831}
 832
 833static int vxlan_opt_to_nlattr(struct sk_buff *skb,
 834			       const void *tun_opts, int swkey_tun_opts_len)
 835{
 836	const struct vxlan_metadata *opts = tun_opts;
 837	struct nlattr *nla;
 838
 839	nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
 840	if (!nla)
 841		return -EMSGSIZE;
 842
 843	if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
 844		return -EMSGSIZE;
 845
 846	nla_nest_end(skb, nla);
 847	return 0;
 848}
 849
 850static int __ip_tun_to_nlattr(struct sk_buff *skb,
 851			      const struct ip_tunnel_key *output,
 852			      const void *tun_opts, int swkey_tun_opts_len,
 853			      unsigned short tun_proto)
 854{
 855	if (output->tun_flags & TUNNEL_KEY &&
 856	    nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
 857			 OVS_TUNNEL_KEY_ATTR_PAD))
 858		return -EMSGSIZE;
 
 
 
 
 
 859	switch (tun_proto) {
 860	case AF_INET:
 861		if (output->u.ipv4.src &&
 862		    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
 863				    output->u.ipv4.src))
 864			return -EMSGSIZE;
 865		if (output->u.ipv4.dst &&
 866		    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
 867				    output->u.ipv4.dst))
 868			return -EMSGSIZE;
 869		break;
 870	case AF_INET6:
 871		if (!ipv6_addr_any(&output->u.ipv6.src) &&
 872		    nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
 873				     &output->u.ipv6.src))
 874			return -EMSGSIZE;
 875		if (!ipv6_addr_any(&output->u.ipv6.dst) &&
 876		    nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
 877				     &output->u.ipv6.dst))
 878			return -EMSGSIZE;
 879		break;
 880	}
 881	if (output->tos &&
 882	    nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
 883		return -EMSGSIZE;
 884	if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
 885		return -EMSGSIZE;
 886	if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
 887	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
 888		return -EMSGSIZE;
 889	if ((output->tun_flags & TUNNEL_CSUM) &&
 890	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
 891		return -EMSGSIZE;
 892	if (output->tp_src &&
 893	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
 894		return -EMSGSIZE;
 895	if (output->tp_dst &&
 896	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
 897		return -EMSGSIZE;
 898	if ((output->tun_flags & TUNNEL_OAM) &&
 899	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
 900		return -EMSGSIZE;
 901	if (swkey_tun_opts_len) {
 902		if (output->tun_flags & TUNNEL_GENEVE_OPT &&
 903		    nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
 904			    swkey_tun_opts_len, tun_opts))
 905			return -EMSGSIZE;
 906		else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
 907			 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
 908			return -EMSGSIZE;
 909		else if (output->tun_flags & TUNNEL_ERSPAN_OPT &&
 910			 nla_put(skb, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS,
 911				 swkey_tun_opts_len, tun_opts))
 912			return -EMSGSIZE;
 913	}
 914
 915	return 0;
 916}
 917
 918static int ip_tun_to_nlattr(struct sk_buff *skb,
 919			    const struct ip_tunnel_key *output,
 920			    const void *tun_opts, int swkey_tun_opts_len,
 921			    unsigned short tun_proto)
 922{
 923	struct nlattr *nla;
 924	int err;
 925
 926	nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
 927	if (!nla)
 928		return -EMSGSIZE;
 929
 930	err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
 931				 tun_proto);
 932	if (err)
 933		return err;
 934
 935	nla_nest_end(skb, nla);
 936	return 0;
 937}
 938
 939int ovs_nla_put_tunnel_info(struct sk_buff *skb,
 940			    struct ip_tunnel_info *tun_info)
 941{
 942	return __ip_tun_to_nlattr(skb, &tun_info->key,
 943				  ip_tunnel_info_opts(tun_info),
 944				  tun_info->options_len,
 945				  ip_tunnel_info_af(tun_info));
 946}
 947
 948static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
 949				    const struct nlattr *a[],
 950				    bool is_mask, bool inner)
 951{
 952	__be16 tci = 0;
 953	__be16 tpid = 0;
 954
 955	if (a[OVS_KEY_ATTR_VLAN])
 956		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
 957
 958	if (a[OVS_KEY_ATTR_ETHERTYPE])
 959		tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
 960
 961	if (likely(!inner)) {
 962		SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
 963		SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
 964	} else {
 965		SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
 966		SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
 967	}
 968	return 0;
 969}
 970
 971static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
 972				      u64 key_attrs, bool inner,
 973				      const struct nlattr **a, bool log)
 974{
 975	__be16 tci = 0;
 976
 977	if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
 978	      (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
 979	       eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
 980		/* Not a VLAN. */
 981		return 0;
 982	}
 983
 984	if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
 985	      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
 986		OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
 987		return -EINVAL;
 988	}
 989
 990	if (a[OVS_KEY_ATTR_VLAN])
 991		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
 992
 993	if (!(tci & htons(VLAN_TAG_PRESENT))) {
 994		if (tci) {
 995			OVS_NLERR(log, "%s TCI does not have VLAN_TAG_PRESENT bit set.",
 996				  (inner) ? "C-VLAN" : "VLAN");
 997			return -EINVAL;
 998		} else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
 999			/* Corner case for truncated VLAN header. */
1000			OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
1001				  (inner) ? "C-VLAN" : "VLAN");
1002			return -EINVAL;
1003		}
1004	}
1005
1006	return 1;
1007}
1008
1009static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1010					   u64 key_attrs, bool inner,
1011					   const struct nlattr **a, bool log)
1012{
1013	__be16 tci = 0;
1014	__be16 tpid = 0;
1015	bool encap_valid = !!(match->key->eth.vlan.tci &
1016			      htons(VLAN_TAG_PRESENT));
1017	bool i_encap_valid = !!(match->key->eth.cvlan.tci &
1018				htons(VLAN_TAG_PRESENT));
1019
1020	if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1021		/* Not a VLAN. */
1022		return 0;
1023	}
1024
1025	if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
1026		OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
1027			  (inner) ? "C-VLAN" : "VLAN");
1028		return -EINVAL;
1029	}
1030
1031	if (a[OVS_KEY_ATTR_VLAN])
1032		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1033
1034	if (a[OVS_KEY_ATTR_ETHERTYPE])
1035		tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1036
1037	if (tpid != htons(0xffff)) {
1038		OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
1039			  (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
1040		return -EINVAL;
1041	}
1042	if (!(tci & htons(VLAN_TAG_PRESENT))) {
1043		OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_TAG_PRESENT bit.",
1044			  (inner) ? "C-VLAN" : "VLAN");
1045		return -EINVAL;
1046	}
1047
1048	return 1;
1049}
1050
1051static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1052				     u64 *key_attrs, bool inner,
1053				     const struct nlattr **a, bool is_mask,
1054				     bool log)
1055{
1056	int err;
1057	const struct nlattr *encap;
1058
1059	if (!is_mask)
1060		err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1061						 a, log);
1062	else
1063		err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1064						      a, log);
1065	if (err <= 0)
1066		return err;
1067
1068	err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1069	if (err)
1070		return err;
1071
1072	*key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1073	*key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1074	*key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1075
1076	encap = a[OVS_KEY_ATTR_ENCAP];
1077
1078	if (!is_mask)
1079		err = parse_flow_nlattrs(encap, a, key_attrs, log);
1080	else
1081		err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1082
1083	return err;
1084}
1085
1086static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1087				   u64 *key_attrs, const struct nlattr **a,
1088				   bool is_mask, bool log)
1089{
1090	int err;
1091	bool encap_valid = false;
1092
1093	err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1094					is_mask, log);
1095	if (err)
1096		return err;
1097
1098	encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_TAG_PRESENT));
1099	if (encap_valid) {
1100		err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1101						is_mask, log);
1102		if (err)
1103			return err;
1104	}
1105
1106	return 0;
1107}
1108
1109static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1110				       u64 *attrs, const struct nlattr **a,
1111				       bool is_mask, bool log)
1112{
1113	__be16 eth_type;
1114
1115	eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1116	if (is_mask) {
1117		/* Always exact match EtherType. */
1118		eth_type = htons(0xffff);
1119	} else if (!eth_proto_is_802_3(eth_type)) {
1120		OVS_NLERR(log, "EtherType %x is less than min %x",
1121				ntohs(eth_type), ETH_P_802_3_MIN);
1122		return -EINVAL;
1123	}
1124
1125	SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1126	*attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1127	return 0;
1128}
1129
1130static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1131				 u64 *attrs, const struct nlattr **a,
1132				 bool is_mask, bool log)
1133{
1134	u8 mac_proto = MAC_PROTO_ETHERNET;
1135
1136	if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1137		u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1138
1139		SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1140		*attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1141	}
1142
1143	if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1144		u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1145
1146		SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1147		*attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1148	}
1149
1150	if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1151		SW_FLOW_KEY_PUT(match, phy.priority,
1152			  nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1153		*attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1154	}
1155
1156	if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1157		u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1158
1159		if (is_mask) {
1160			in_port = 0xffffffff; /* Always exact match in_port. */
1161		} else if (in_port >= DP_MAX_PORTS) {
1162			OVS_NLERR(log, "Port %d exceeds max allowable %d",
1163				  in_port, DP_MAX_PORTS);
1164			return -EINVAL;
1165		}
1166
1167		SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1168		*attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1169	} else if (!is_mask) {
1170		SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1171	}
1172
1173	if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1174		uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1175
1176		SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1177		*attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1178	}
1179	if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1180		if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1181				       is_mask, log) < 0)
1182			return -EINVAL;
1183		*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1184	}
1185
1186	if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1187	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1188		u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1189
1190		if (ct_state & ~CT_SUPPORTED_MASK) {
1191			OVS_NLERR(log, "ct_state flags %08x unsupported",
1192				  ct_state);
1193			return -EINVAL;
1194		}
1195
1196		SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1197		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1198	}
1199	if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1200	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1201		u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1202
1203		SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1204		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1205	}
1206	if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1207	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1208		u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1209
1210		SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1211		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1212	}
1213	if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1214	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1215		const struct ovs_key_ct_labels *cl;
1216
1217		cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1218		SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1219				   sizeof(*cl), is_mask);
1220		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1221	}
1222	if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1223		const struct ovs_key_ct_tuple_ipv4 *ct;
1224
1225		ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1226
1227		SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1228		SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1229		SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1230		SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1231		SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1232		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1233	}
1234	if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1235		const struct ovs_key_ct_tuple_ipv6 *ct;
1236
1237		ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1238
1239		SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1240				   sizeof(match->key->ipv6.ct_orig.src),
1241				   is_mask);
1242		SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1243				   sizeof(match->key->ipv6.ct_orig.dst),
1244				   is_mask);
1245		SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1246		SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1247		SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1248		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1249	}
1250
1251	/* For layer 3 packets the Ethernet type is provided
1252	 * and treated as metadata but no MAC addresses are provided.
1253	 */
1254	if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1255	    (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1256		mac_proto = MAC_PROTO_NONE;
1257
1258	/* Always exact match mac_proto */
1259	SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1260
1261	if (mac_proto == MAC_PROTO_NONE)
1262		return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1263						   log);
1264
1265	return 0;
1266}
1267
1268int nsh_hdr_from_nlattr(const struct nlattr *attr,
1269			struct nshhdr *nh, size_t size)
1270{
1271	struct nlattr *a;
1272	int rem;
1273	u8 flags = 0;
1274	u8 ttl = 0;
1275	int mdlen = 0;
1276
1277	/* validate_nsh has check this, so we needn't do duplicate check here
1278	 */
1279	if (size < NSH_BASE_HDR_LEN)
1280		return -ENOBUFS;
1281
1282	nla_for_each_nested(a, attr, rem) {
1283		int type = nla_type(a);
1284
1285		switch (type) {
1286		case OVS_NSH_KEY_ATTR_BASE: {
1287			const struct ovs_nsh_key_base *base = nla_data(a);
1288
1289			flags = base->flags;
1290			ttl = base->ttl;
1291			nh->np = base->np;
1292			nh->mdtype = base->mdtype;
1293			nh->path_hdr = base->path_hdr;
1294			break;
1295		}
1296		case OVS_NSH_KEY_ATTR_MD1:
1297			mdlen = nla_len(a);
1298			if (mdlen > size - NSH_BASE_HDR_LEN)
1299				return -ENOBUFS;
1300			memcpy(&nh->md1, nla_data(a), mdlen);
1301			break;
1302
1303		case OVS_NSH_KEY_ATTR_MD2:
1304			mdlen = nla_len(a);
1305			if (mdlen > size - NSH_BASE_HDR_LEN)
1306				return -ENOBUFS;
1307			memcpy(&nh->md2, nla_data(a), mdlen);
1308			break;
1309
1310		default:
1311			return -EINVAL;
1312		}
1313	}
1314
1315	/* nsh header length  = NSH_BASE_HDR_LEN + mdlen */
1316	nh->ver_flags_ttl_len = 0;
1317	nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1318
1319	return 0;
1320}
1321
1322int nsh_key_from_nlattr(const struct nlattr *attr,
1323			struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1324{
1325	struct nlattr *a;
1326	int rem;
1327
1328	/* validate_nsh has check this, so we needn't do duplicate check here
1329	 */
1330	nla_for_each_nested(a, attr, rem) {
1331		int type = nla_type(a);
1332
1333		switch (type) {
1334		case OVS_NSH_KEY_ATTR_BASE: {
1335			const struct ovs_nsh_key_base *base = nla_data(a);
1336			const struct ovs_nsh_key_base *base_mask = base + 1;
1337
1338			nsh->base = *base;
1339			nsh_mask->base = *base_mask;
1340			break;
1341		}
1342		case OVS_NSH_KEY_ATTR_MD1: {
1343			const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1344			const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1345
1346			memcpy(nsh->context, md1->context, sizeof(*md1));
1347			memcpy(nsh_mask->context, md1_mask->context,
1348			       sizeof(*md1_mask));
1349			break;
1350		}
1351		case OVS_NSH_KEY_ATTR_MD2:
1352			/* Not supported yet */
1353			return -ENOTSUPP;
1354		default:
1355			return -EINVAL;
1356		}
1357	}
1358
1359	return 0;
1360}
1361
1362static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1363				   struct sw_flow_match *match, bool is_mask,
1364				   bool is_push_nsh, bool log)
1365{
1366	struct nlattr *a;
1367	int rem;
1368	bool has_base = false;
1369	bool has_md1 = false;
1370	bool has_md2 = false;
1371	u8 mdtype = 0;
1372	int mdlen = 0;
1373
1374	if (WARN_ON(is_push_nsh && is_mask))
1375		return -EINVAL;
1376
1377	nla_for_each_nested(a, attr, rem) {
1378		int type = nla_type(a);
1379		int i;
1380
1381		if (type > OVS_NSH_KEY_ATTR_MAX) {
1382			OVS_NLERR(log, "nsh attr %d is out of range max %d",
1383				  type, OVS_NSH_KEY_ATTR_MAX);
1384			return -EINVAL;
1385		}
1386
1387		if (!check_attr_len(nla_len(a),
1388				    ovs_nsh_key_attr_lens[type].len)) {
1389			OVS_NLERR(
1390			    log,
1391			    "nsh attr %d has unexpected len %d expected %d",
1392			    type,
1393			    nla_len(a),
1394			    ovs_nsh_key_attr_lens[type].len
1395			);
1396			return -EINVAL;
1397		}
1398
1399		switch (type) {
1400		case OVS_NSH_KEY_ATTR_BASE: {
1401			const struct ovs_nsh_key_base *base = nla_data(a);
1402
1403			has_base = true;
1404			mdtype = base->mdtype;
1405			SW_FLOW_KEY_PUT(match, nsh.base.flags,
1406					base->flags, is_mask);
1407			SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1408					base->ttl, is_mask);
1409			SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1410					base->mdtype, is_mask);
1411			SW_FLOW_KEY_PUT(match, nsh.base.np,
1412					base->np, is_mask);
1413			SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1414					base->path_hdr, is_mask);
1415			break;
1416		}
1417		case OVS_NSH_KEY_ATTR_MD1: {
1418			const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1419
1420			has_md1 = true;
1421			for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1422				SW_FLOW_KEY_PUT(match, nsh.context[i],
1423						md1->context[i], is_mask);
1424			break;
1425		}
1426		case OVS_NSH_KEY_ATTR_MD2:
1427			if (!is_push_nsh) /* Not supported MD type 2 yet */
1428				return -ENOTSUPP;
1429
1430			has_md2 = true;
1431			mdlen = nla_len(a);
1432			if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1433				OVS_NLERR(
1434				    log,
1435				    "Invalid MD length %d for MD type %d",
1436				    mdlen,
1437				    mdtype
1438				);
1439				return -EINVAL;
1440			}
1441			break;
1442		default:
1443			OVS_NLERR(log, "Unknown nsh attribute %d",
1444				  type);
1445			return -EINVAL;
1446		}
1447	}
1448
1449	if (rem > 0) {
1450		OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1451		return -EINVAL;
1452	}
1453
1454	if (has_md1 && has_md2) {
1455		OVS_NLERR(
1456		    1,
1457		    "invalid nsh attribute: md1 and md2 are exclusive."
1458		);
1459		return -EINVAL;
1460	}
1461
1462	if (!is_mask) {
1463		if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1464		    (has_md2 && mdtype != NSH_M_TYPE2)) {
1465			OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1466				  mdtype);
1467			return -EINVAL;
1468		}
1469
1470		if (is_push_nsh &&
1471		    (!has_base || (!has_md1 && !has_md2))) {
1472			OVS_NLERR(
1473			    1,
1474			    "push_nsh: missing base or metadata attributes"
1475			);
1476			return -EINVAL;
1477		}
1478	}
1479
1480	return 0;
1481}
1482
1483static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1484				u64 attrs, const struct nlattr **a,
1485				bool is_mask, bool log)
1486{
1487	int err;
1488
1489	err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1490	if (err)
1491		return err;
1492
1493	if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1494		const struct ovs_key_ethernet *eth_key;
1495
1496		eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1497		SW_FLOW_KEY_MEMCPY(match, eth.src,
1498				eth_key->eth_src, ETH_ALEN, is_mask);
1499		SW_FLOW_KEY_MEMCPY(match, eth.dst,
1500				eth_key->eth_dst, ETH_ALEN, is_mask);
1501		attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1502
1503		if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1504			/* VLAN attribute is always parsed before getting here since it
1505			 * may occur multiple times.
1506			 */
1507			OVS_NLERR(log, "VLAN attribute unexpected.");
1508			return -EINVAL;
1509		}
1510
1511		if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1512			err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1513							  log);
1514			if (err)
1515				return err;
1516		} else if (!is_mask) {
1517			SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1518		}
1519	} else if (!match->key->eth.type) {
1520		OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1521		return -EINVAL;
1522	}
1523
1524	if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1525		const struct ovs_key_ipv4 *ipv4_key;
1526
1527		ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1528		if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1529			OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1530				  ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1531			return -EINVAL;
1532		}
1533		SW_FLOW_KEY_PUT(match, ip.proto,
1534				ipv4_key->ipv4_proto, is_mask);
1535		SW_FLOW_KEY_PUT(match, ip.tos,
1536				ipv4_key->ipv4_tos, is_mask);
1537		SW_FLOW_KEY_PUT(match, ip.ttl,
1538				ipv4_key->ipv4_ttl, is_mask);
1539		SW_FLOW_KEY_PUT(match, ip.frag,
1540				ipv4_key->ipv4_frag, is_mask);
1541		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1542				ipv4_key->ipv4_src, is_mask);
1543		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1544				ipv4_key->ipv4_dst, is_mask);
1545		attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1546	}
1547
1548	if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1549		const struct ovs_key_ipv6 *ipv6_key;
1550
1551		ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1552		if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1553			OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1554				  ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1555			return -EINVAL;
1556		}
1557
1558		if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1559			OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1560				  ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1561			return -EINVAL;
1562		}
1563
1564		SW_FLOW_KEY_PUT(match, ipv6.label,
1565				ipv6_key->ipv6_label, is_mask);
1566		SW_FLOW_KEY_PUT(match, ip.proto,
1567				ipv6_key->ipv6_proto, is_mask);
1568		SW_FLOW_KEY_PUT(match, ip.tos,
1569				ipv6_key->ipv6_tclass, is_mask);
1570		SW_FLOW_KEY_PUT(match, ip.ttl,
1571				ipv6_key->ipv6_hlimit, is_mask);
1572		SW_FLOW_KEY_PUT(match, ip.frag,
1573				ipv6_key->ipv6_frag, is_mask);
1574		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1575				ipv6_key->ipv6_src,
1576				sizeof(match->key->ipv6.addr.src),
1577				is_mask);
1578		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1579				ipv6_key->ipv6_dst,
1580				sizeof(match->key->ipv6.addr.dst),
1581				is_mask);
1582
1583		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1584	}
1585
1586	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1587		const struct ovs_key_arp *arp_key;
1588
1589		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1590		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1591			OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1592				  arp_key->arp_op);
1593			return -EINVAL;
1594		}
1595
1596		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1597				arp_key->arp_sip, is_mask);
1598		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1599			arp_key->arp_tip, is_mask);
1600		SW_FLOW_KEY_PUT(match, ip.proto,
1601				ntohs(arp_key->arp_op), is_mask);
1602		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1603				arp_key->arp_sha, ETH_ALEN, is_mask);
1604		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1605				arp_key->arp_tha, ETH_ALEN, is_mask);
1606
1607		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1608	}
1609
1610	if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1611		if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1612					    is_mask, false, log) < 0)
1613			return -EINVAL;
1614		attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1615	}
1616
1617	if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1618		const struct ovs_key_mpls *mpls_key;
 
 
1619
1620		mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1621		SW_FLOW_KEY_PUT(match, mpls.top_lse,
1622				mpls_key->mpls_lse, is_mask);
 
 
 
 
 
 
 
 
 
 
 
 
 
1623
1624		attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1625	 }
1626
1627	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1628		const struct ovs_key_tcp *tcp_key;
1629
1630		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1631		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1632		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1633		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1634	}
1635
1636	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1637		SW_FLOW_KEY_PUT(match, tp.flags,
1638				nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1639				is_mask);
1640		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1641	}
1642
1643	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1644		const struct ovs_key_udp *udp_key;
1645
1646		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1647		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1648		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1649		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1650	}
1651
1652	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1653		const struct ovs_key_sctp *sctp_key;
1654
1655		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1656		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1657		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1658		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1659	}
1660
1661	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1662		const struct ovs_key_icmp *icmp_key;
1663
1664		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1665		SW_FLOW_KEY_PUT(match, tp.src,
1666				htons(icmp_key->icmp_type), is_mask);
1667		SW_FLOW_KEY_PUT(match, tp.dst,
1668				htons(icmp_key->icmp_code), is_mask);
1669		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1670	}
1671
1672	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1673		const struct ovs_key_icmpv6 *icmpv6_key;
1674
1675		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1676		SW_FLOW_KEY_PUT(match, tp.src,
1677				htons(icmpv6_key->icmpv6_type), is_mask);
1678		SW_FLOW_KEY_PUT(match, tp.dst,
1679				htons(icmpv6_key->icmpv6_code), is_mask);
1680		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1681	}
1682
1683	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1684		const struct ovs_key_nd *nd_key;
1685
1686		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1687		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1688			nd_key->nd_target,
1689			sizeof(match->key->ipv6.nd.target),
1690			is_mask);
1691		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1692			nd_key->nd_sll, ETH_ALEN, is_mask);
1693		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1694				nd_key->nd_tll, ETH_ALEN, is_mask);
1695		attrs &= ~(1 << OVS_KEY_ATTR_ND);
1696	}
1697
1698	if (attrs != 0) {
1699		OVS_NLERR(log, "Unknown key attributes %llx",
1700			  (unsigned long long)attrs);
1701		return -EINVAL;
1702	}
1703
1704	return 0;
1705}
1706
1707static void nlattr_set(struct nlattr *attr, u8 val,
1708		       const struct ovs_len_tbl *tbl)
1709{
1710	struct nlattr *nla;
1711	int rem;
1712
1713	/* The nlattr stream should already have been validated */
1714	nla_for_each_nested(nla, attr, rem) {
1715		if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1716			nlattr_set(nla, val, tbl[nla_type(nla)].next ? : tbl);
1717		else
1718			memset(nla_data(nla), val, nla_len(nla));
1719
1720		if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1721			*(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1722	}
1723}
1724
1725static void mask_set_nlattr(struct nlattr *attr, u8 val)
1726{
1727	nlattr_set(attr, val, ovs_key_lens);
1728}
1729
1730/**
1731 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1732 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1733 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1734 * does not include any don't care bit.
1735 * @net: Used to determine per-namespace field support.
1736 * @match: receives the extracted flow match information.
1737 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1738 * sequence. The fields should of the packet that triggered the creation
1739 * of this flow.
1740 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1741 * attribute specifies the mask field of the wildcarded flow.
1742 * @log: Boolean to allow kernel error logging.  Normally true, but when
1743 * probing for feature compatibility this should be passed in as false to
1744 * suppress unnecessary error logging.
1745 */
1746int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1747		      const struct nlattr *nla_key,
1748		      const struct nlattr *nla_mask,
1749		      bool log)
1750{
1751	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1752	struct nlattr *newmask = NULL;
1753	u64 key_attrs = 0;
1754	u64 mask_attrs = 0;
1755	int err;
1756
1757	err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1758	if (err)
1759		return err;
1760
1761	err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1762	if (err)
1763		return err;
1764
1765	err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1766	if (err)
1767		return err;
1768
1769	if (match->mask) {
1770		if (!nla_mask) {
1771			/* Create an exact match mask. We need to set to 0xff
1772			 * all the 'match->mask' fields that have been touched
1773			 * in 'match->key'. We cannot simply memset
1774			 * 'match->mask', because padding bytes and fields not
1775			 * specified in 'match->key' should be left to 0.
1776			 * Instead, we use a stream of netlink attributes,
1777			 * copied from 'key' and set to 0xff.
1778			 * ovs_key_from_nlattrs() will take care of filling
1779			 * 'match->mask' appropriately.
1780			 */
1781			newmask = kmemdup(nla_key,
1782					  nla_total_size(nla_len(nla_key)),
1783					  GFP_KERNEL);
1784			if (!newmask)
1785				return -ENOMEM;
1786
1787			mask_set_nlattr(newmask, 0xff);
1788
1789			/* The userspace does not send tunnel attributes that
1790			 * are 0, but we should not wildcard them nonetheless.
1791			 */
1792			if (match->key->tun_proto)
1793				SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1794							 0xff, true);
1795
1796			nla_mask = newmask;
1797		}
1798
1799		err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1800		if (err)
1801			goto free_newmask;
1802
1803		/* Always match on tci. */
1804		SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1805		SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1806
1807		err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1808		if (err)
1809			goto free_newmask;
1810
1811		err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1812					   log);
1813		if (err)
1814			goto free_newmask;
1815	}
1816
1817	if (!match_validate(match, key_attrs, mask_attrs, log))
1818		err = -EINVAL;
1819
1820free_newmask:
1821	kfree(newmask);
1822	return err;
1823}
1824
1825static size_t get_ufid_len(const struct nlattr *attr, bool log)
1826{
1827	size_t len;
1828
1829	if (!attr)
1830		return 0;
1831
1832	len = nla_len(attr);
1833	if (len < 1 || len > MAX_UFID_LENGTH) {
1834		OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1835			  nla_len(attr), MAX_UFID_LENGTH);
1836		return 0;
1837	}
1838
1839	return len;
1840}
1841
1842/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1843 * or false otherwise.
1844 */
1845bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1846		      bool log)
1847{
1848	sfid->ufid_len = get_ufid_len(attr, log);
1849	if (sfid->ufid_len)
1850		memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1851
1852	return sfid->ufid_len;
1853}
1854
1855int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1856			   const struct sw_flow_key *key, bool log)
1857{
1858	struct sw_flow_key *new_key;
1859
1860	if (ovs_nla_get_ufid(sfid, ufid, log))
1861		return 0;
1862
1863	/* If UFID was not provided, use unmasked key. */
1864	new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1865	if (!new_key)
1866		return -ENOMEM;
1867	memcpy(new_key, key, sizeof(*key));
1868	sfid->unmasked_key = new_key;
1869
1870	return 0;
1871}
1872
1873u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1874{
1875	return attr ? nla_get_u32(attr) : 0;
1876}
1877
1878/**
1879 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1880 * @net: Network namespace.
1881 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1882 * metadata.
1883 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1884 * attributes.
1885 * @attrs: Bit mask for the netlink attributes included in @a.
1886 * @log: Boolean to allow kernel error logging.  Normally true, but when
1887 * probing for feature compatibility this should be passed in as false to
1888 * suppress unnecessary error logging.
1889 *
1890 * This parses a series of Netlink attributes that form a flow key, which must
1891 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1892 * get the metadata, that is, the parts of the flow key that cannot be
1893 * extracted from the packet itself.
1894 *
1895 * This must be called before the packet key fields are filled in 'key'.
1896 */
1897
1898int ovs_nla_get_flow_metadata(struct net *net,
1899			      const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1900			      u64 attrs, struct sw_flow_key *key, bool log)
1901{
1902	struct sw_flow_match match;
1903
1904	memset(&match, 0, sizeof(match));
1905	match.key = key;
1906
1907	key->ct_state = 0;
1908	key->ct_zone = 0;
1909	key->ct_orig_proto = 0;
1910	memset(&key->ct, 0, sizeof(key->ct));
1911	memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1912	memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1913
1914	key->phy.in_port = DP_MAX_PORTS;
1915
1916	return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1917}
1918
1919static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1920			    bool is_mask)
1921{
1922	__be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1923
1924	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1925	    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1926		return -EMSGSIZE;
1927	return 0;
1928}
1929
1930static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1931			     struct sk_buff *skb)
1932{
1933	struct nlattr *start;
1934
1935	start = nla_nest_start(skb, OVS_KEY_ATTR_NSH);
1936	if (!start)
1937		return -EMSGSIZE;
1938
1939	if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
1940		goto nla_put_failure;
1941
1942	if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
1943		if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
1944			    sizeof(nsh->context), nsh->context))
1945			goto nla_put_failure;
1946	}
1947
1948	/* Don't support MD type 2 yet */
1949
1950	nla_nest_end(skb, start);
1951
1952	return 0;
1953
1954nla_put_failure:
1955	return -EMSGSIZE;
1956}
1957
1958static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1959			     const struct sw_flow_key *output, bool is_mask,
1960			     struct sk_buff *skb)
1961{
1962	struct ovs_key_ethernet *eth_key;
1963	struct nlattr *nla;
1964	struct nlattr *encap = NULL;
1965	struct nlattr *in_encap = NULL;
1966
1967	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1968		goto nla_put_failure;
1969
1970	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1971		goto nla_put_failure;
1972
1973	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1974		goto nla_put_failure;
1975
1976	if ((swkey->tun_proto || is_mask)) {
1977		const void *opts = NULL;
1978
1979		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1980			opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1981
1982		if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1983				     swkey->tun_opts_len, swkey->tun_proto))
1984			goto nla_put_failure;
1985	}
1986
1987	if (swkey->phy.in_port == DP_MAX_PORTS) {
1988		if (is_mask && (output->phy.in_port == 0xffff))
1989			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1990				goto nla_put_failure;
1991	} else {
1992		u16 upper_u16;
1993		upper_u16 = !is_mask ? 0 : 0xffff;
1994
1995		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1996				(upper_u16 << 16) | output->phy.in_port))
1997			goto nla_put_failure;
1998	}
1999
2000	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2001		goto nla_put_failure;
2002
2003	if (ovs_ct_put_key(swkey, output, skb))
2004		goto nla_put_failure;
2005
2006	if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2007		nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2008		if (!nla)
2009			goto nla_put_failure;
2010
2011		eth_key = nla_data(nla);
2012		ether_addr_copy(eth_key->eth_src, output->eth.src);
2013		ether_addr_copy(eth_key->eth_dst, output->eth.dst);
2014
2015		if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
2016			if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
2017				goto nla_put_failure;
2018			encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
2019			if (!swkey->eth.vlan.tci)
2020				goto unencap;
2021
2022			if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
2023				if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
2024					goto nla_put_failure;
2025				in_encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
 
2026				if (!swkey->eth.cvlan.tci)
2027					goto unencap;
2028			}
2029		}
2030
2031		if (swkey->eth.type == htons(ETH_P_802_2)) {
2032			/*
2033			* Ethertype 802.2 is represented in the netlink with omitted
2034			* OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2035			* 0xffff in the mask attribute.  Ethertype can also
2036			* be wildcarded.
2037			*/
2038			if (is_mask && output->eth.type)
2039				if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2040							output->eth.type))
2041					goto nla_put_failure;
2042			goto unencap;
2043		}
2044	}
2045
2046	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2047		goto nla_put_failure;
2048
2049	if (eth_type_vlan(swkey->eth.type)) {
2050		/* There are 3 VLAN tags, we don't know anything about the rest
2051		 * of the packet, so truncate here.
2052		 */
2053		WARN_ON_ONCE(!(encap && in_encap));
2054		goto unencap;
2055	}
2056
2057	if (swkey->eth.type == htons(ETH_P_IP)) {
2058		struct ovs_key_ipv4 *ipv4_key;
2059
2060		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2061		if (!nla)
2062			goto nla_put_failure;
2063		ipv4_key = nla_data(nla);
2064		ipv4_key->ipv4_src = output->ipv4.addr.src;
2065		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2066		ipv4_key->ipv4_proto = output->ip.proto;
2067		ipv4_key->ipv4_tos = output->ip.tos;
2068		ipv4_key->ipv4_ttl = output->ip.ttl;
2069		ipv4_key->ipv4_frag = output->ip.frag;
2070	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2071		struct ovs_key_ipv6 *ipv6_key;
2072
2073		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2074		if (!nla)
2075			goto nla_put_failure;
2076		ipv6_key = nla_data(nla);
2077		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2078				sizeof(ipv6_key->ipv6_src));
2079		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2080				sizeof(ipv6_key->ipv6_dst));
2081		ipv6_key->ipv6_label = output->ipv6.label;
2082		ipv6_key->ipv6_proto = output->ip.proto;
2083		ipv6_key->ipv6_tclass = output->ip.tos;
2084		ipv6_key->ipv6_hlimit = output->ip.ttl;
2085		ipv6_key->ipv6_frag = output->ip.frag;
2086	} else if (swkey->eth.type == htons(ETH_P_NSH)) {
2087		if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2088			goto nla_put_failure;
2089	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
2090		   swkey->eth.type == htons(ETH_P_RARP)) {
2091		struct ovs_key_arp *arp_key;
2092
2093		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2094		if (!nla)
2095			goto nla_put_failure;
2096		arp_key = nla_data(nla);
2097		memset(arp_key, 0, sizeof(struct ovs_key_arp));
2098		arp_key->arp_sip = output->ipv4.addr.src;
2099		arp_key->arp_tip = output->ipv4.addr.dst;
2100		arp_key->arp_op = htons(output->ip.proto);
2101		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2102		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2103	} else if (eth_p_mpls(swkey->eth.type)) {
 
2104		struct ovs_key_mpls *mpls_key;
2105
2106		nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
 
 
2107		if (!nla)
2108			goto nla_put_failure;
 
2109		mpls_key = nla_data(nla);
2110		mpls_key->mpls_lse = output->mpls.top_lse;
 
2111	}
2112
2113	if ((swkey->eth.type == htons(ETH_P_IP) ||
2114	     swkey->eth.type == htons(ETH_P_IPV6)) &&
2115	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2116
2117		if (swkey->ip.proto == IPPROTO_TCP) {
2118			struct ovs_key_tcp *tcp_key;
2119
2120			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2121			if (!nla)
2122				goto nla_put_failure;
2123			tcp_key = nla_data(nla);
2124			tcp_key->tcp_src = output->tp.src;
2125			tcp_key->tcp_dst = output->tp.dst;
2126			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2127					 output->tp.flags))
2128				goto nla_put_failure;
2129		} else if (swkey->ip.proto == IPPROTO_UDP) {
2130			struct ovs_key_udp *udp_key;
2131
2132			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2133			if (!nla)
2134				goto nla_put_failure;
2135			udp_key = nla_data(nla);
2136			udp_key->udp_src = output->tp.src;
2137			udp_key->udp_dst = output->tp.dst;
2138		} else if (swkey->ip.proto == IPPROTO_SCTP) {
2139			struct ovs_key_sctp *sctp_key;
2140
2141			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2142			if (!nla)
2143				goto nla_put_failure;
2144			sctp_key = nla_data(nla);
2145			sctp_key->sctp_src = output->tp.src;
2146			sctp_key->sctp_dst = output->tp.dst;
2147		} else if (swkey->eth.type == htons(ETH_P_IP) &&
2148			   swkey->ip.proto == IPPROTO_ICMP) {
2149			struct ovs_key_icmp *icmp_key;
2150
2151			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2152			if (!nla)
2153				goto nla_put_failure;
2154			icmp_key = nla_data(nla);
2155			icmp_key->icmp_type = ntohs(output->tp.src);
2156			icmp_key->icmp_code = ntohs(output->tp.dst);
2157		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2158			   swkey->ip.proto == IPPROTO_ICMPV6) {
2159			struct ovs_key_icmpv6 *icmpv6_key;
2160
2161			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2162						sizeof(*icmpv6_key));
2163			if (!nla)
2164				goto nla_put_failure;
2165			icmpv6_key = nla_data(nla);
2166			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2167			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2168
2169			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
2170			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
2171				struct ovs_key_nd *nd_key;
2172
2173				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2174				if (!nla)
2175					goto nla_put_failure;
2176				nd_key = nla_data(nla);
2177				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2178							sizeof(nd_key->nd_target));
2179				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2180				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2181			}
2182		}
2183	}
2184
2185unencap:
2186	if (in_encap)
2187		nla_nest_end(skb, in_encap);
2188	if (encap)
2189		nla_nest_end(skb, encap);
2190
2191	return 0;
2192
2193nla_put_failure:
2194	return -EMSGSIZE;
2195}
2196
2197int ovs_nla_put_key(const struct sw_flow_key *swkey,
2198		    const struct sw_flow_key *output, int attr, bool is_mask,
2199		    struct sk_buff *skb)
2200{
2201	int err;
2202	struct nlattr *nla;
2203
2204	nla = nla_nest_start(skb, attr);
2205	if (!nla)
2206		return -EMSGSIZE;
2207	err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2208	if (err)
2209		return err;
2210	nla_nest_end(skb, nla);
2211
2212	return 0;
2213}
2214
2215/* Called with ovs_mutex or RCU read lock. */
2216int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2217{
2218	if (ovs_identifier_is_ufid(&flow->id))
2219		return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2220			       flow->id.ufid);
2221
2222	return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2223			       OVS_FLOW_ATTR_KEY, false, skb);
2224}
2225
2226/* Called with ovs_mutex or RCU read lock. */
2227int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2228{
2229	return ovs_nla_put_key(&flow->key, &flow->key,
2230				OVS_FLOW_ATTR_KEY, false, skb);
2231}
2232
2233/* Called with ovs_mutex or RCU read lock. */
2234int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2235{
2236	return ovs_nla_put_key(&flow->key, &flow->mask->key,
2237				OVS_FLOW_ATTR_MASK, true, skb);
2238}
2239
2240#define MAX_ACTIONS_BUFSIZE	(32 * 1024)
2241
2242static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2243{
2244	struct sw_flow_actions *sfa;
2245
2246	WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2247
2248	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
2249	if (!sfa)
2250		return ERR_PTR(-ENOMEM);
2251
2252	sfa->actions_len = 0;
2253	return sfa;
2254}
2255
2256static void ovs_nla_free_set_action(const struct nlattr *a)
2257{
2258	const struct nlattr *ovs_key = nla_data(a);
2259	struct ovs_tunnel_info *ovs_tun;
2260
2261	switch (nla_type(ovs_key)) {
2262	case OVS_KEY_ATTR_TUNNEL_INFO:
2263		ovs_tun = nla_data(ovs_key);
2264		dst_release((struct dst_entry *)ovs_tun->tun_dst);
2265		break;
2266	}
2267}
2268
2269void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2270{
2271	const struct nlattr *a;
2272	int rem;
2273
2274	if (!sf_acts)
2275		return;
2276
2277	nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
2278		switch (nla_type(a)) {
2279		case OVS_ACTION_ATTR_SET:
2280			ovs_nla_free_set_action(a);
2281			break;
2282		case OVS_ACTION_ATTR_CT:
2283			ovs_ct_free_action(a);
2284			break;
2285		}
2286	}
2287
2288	kfree(sf_acts);
2289}
2290
2291static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2292{
2293	ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2294}
2295
2296/* Schedules 'sf_acts' to be freed after the next RCU grace period.
2297 * The caller must hold rcu_read_lock for this to be sensible. */
2298void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2299{
2300	call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2301}
2302
2303static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2304				       int attr_len, bool log)
2305{
2306
2307	struct sw_flow_actions *acts;
2308	int new_acts_size;
2309	int req_size = NLA_ALIGN(attr_len);
2310	int next_offset = offsetof(struct sw_flow_actions, actions) +
2311					(*sfa)->actions_len;
2312
2313	if (req_size <= (ksize(*sfa) - next_offset))
2314		goto out;
2315
2316	new_acts_size = ksize(*sfa) * 2;
2317
2318	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2319		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size) {
2320			OVS_NLERR(log, "Flow action size exceeds max %u",
2321				  MAX_ACTIONS_BUFSIZE);
2322			return ERR_PTR(-EMSGSIZE);
2323		}
2324		new_acts_size = MAX_ACTIONS_BUFSIZE;
2325	}
2326
2327	acts = nla_alloc_flow_actions(new_acts_size);
2328	if (IS_ERR(acts))
2329		return (void *)acts;
2330
2331	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2332	acts->actions_len = (*sfa)->actions_len;
2333	acts->orig_len = (*sfa)->orig_len;
2334	kfree(*sfa);
2335	*sfa = acts;
2336
2337out:
2338	(*sfa)->actions_len += req_size;
2339	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2340}
2341
2342static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2343				   int attrtype, void *data, int len, bool log)
2344{
2345	struct nlattr *a;
2346
2347	a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2348	if (IS_ERR(a))
2349		return a;
2350
2351	a->nla_type = attrtype;
2352	a->nla_len = nla_attr_size(len);
2353
2354	if (data)
2355		memcpy(nla_data(a), data, len);
2356	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2357
2358	return a;
2359}
2360
2361int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2362		       int len, bool log)
2363{
2364	struct nlattr *a;
2365
2366	a = __add_action(sfa, attrtype, data, len, log);
2367
2368	return PTR_ERR_OR_ZERO(a);
2369}
2370
2371static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2372					  int attrtype, bool log)
2373{
2374	int used = (*sfa)->actions_len;
2375	int err;
2376
2377	err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2378	if (err)
2379		return err;
2380
2381	return used;
2382}
2383
2384static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2385					 int st_offset)
2386{
2387	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2388							       st_offset);
2389
2390	a->nla_len = sfa->actions_len - st_offset;
2391}
2392
2393static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2394				  const struct sw_flow_key *key,
2395				  struct sw_flow_actions **sfa,
2396				  __be16 eth_type, __be16 vlan_tci, bool log);
 
2397
2398static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2399				    const struct sw_flow_key *key,
2400				    struct sw_flow_actions **sfa,
2401				    __be16 eth_type, __be16 vlan_tci,
2402				    bool log, bool last)
2403{
2404	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2405	const struct nlattr *probability, *actions;
2406	const struct nlattr *a;
2407	int rem, start, err;
2408	struct sample_arg arg;
2409
2410	memset(attrs, 0, sizeof(attrs));
2411	nla_for_each_nested(a, attr, rem) {
2412		int type = nla_type(a);
2413		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2414			return -EINVAL;
2415		attrs[type] = a;
2416	}
2417	if (rem)
2418		return -EINVAL;
2419
2420	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2421	if (!probability || nla_len(probability) != sizeof(u32))
2422		return -EINVAL;
2423
2424	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2425	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2426		return -EINVAL;
2427
2428	/* validation done, copy sample action. */
2429	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2430	if (start < 0)
2431		return start;
2432
2433	/* When both skb and flow may be changed, put the sample
2434	 * into a deferred fifo. On the other hand, if only skb
2435	 * may be modified, the actions can be executed in place.
2436	 *
2437	 * Do this analysis at the flow installation time.
2438	 * Set 'clone_action->exec' to true if the actions can be
2439	 * executed without being deferred.
2440	 *
2441	 * If the sample is the last action, it can always be excuted
2442	 * rather than deferred.
2443	 */
2444	arg.exec = last || !actions_may_change_flow(actions);
2445	arg.probability = nla_get_u32(probability);
2446
2447	err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2448				 log);
2449	if (err)
2450		return err;
2451
2452	err = __ovs_nla_copy_actions(net, actions, key, sfa,
2453				     eth_type, vlan_tci, log);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2454
 
 
2455	if (err)
2456		return err;
2457
2458	add_nested_action_end(*sfa, start);
2459
2460	return 0;
2461}
2462
2463void ovs_match_init(struct sw_flow_match *match,
2464		    struct sw_flow_key *key,
2465		    bool reset_key,
2466		    struct sw_flow_mask *mask)
2467{
2468	memset(match, 0, sizeof(*match));
2469	match->key = key;
2470	match->mask = mask;
2471
2472	if (reset_key)
2473		memset(key, 0, sizeof(*key));
2474
2475	if (mask) {
2476		memset(&mask->key, 0, sizeof(mask->key));
2477		mask->range.start = mask->range.end = 0;
2478	}
2479}
2480
2481static int validate_geneve_opts(struct sw_flow_key *key)
2482{
2483	struct geneve_opt *option;
2484	int opts_len = key->tun_opts_len;
2485	bool crit_opt = false;
2486
2487	option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2488	while (opts_len > 0) {
2489		int len;
2490
2491		if (opts_len < sizeof(*option))
2492			return -EINVAL;
2493
2494		len = sizeof(*option) + option->length * 4;
2495		if (len > opts_len)
2496			return -EINVAL;
2497
2498		crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2499
2500		option = (struct geneve_opt *)((u8 *)option + len);
2501		opts_len -= len;
2502	}
2503
2504	key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2505
2506	return 0;
2507}
2508
2509static int validate_and_copy_set_tun(const struct nlattr *attr,
2510				     struct sw_flow_actions **sfa, bool log)
2511{
2512	struct sw_flow_match match;
2513	struct sw_flow_key key;
2514	struct metadata_dst *tun_dst;
2515	struct ip_tunnel_info *tun_info;
2516	struct ovs_tunnel_info *ovs_tun;
2517	struct nlattr *a;
2518	int err = 0, start, opts_type;
 
2519
 
2520	ovs_match_init(&match, &key, true, NULL);
2521	opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2522	if (opts_type < 0)
2523		return opts_type;
2524
2525	if (key.tun_opts_len) {
2526		switch (opts_type) {
2527		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2528			err = validate_geneve_opts(&key);
2529			if (err < 0)
2530				return err;
 
2531			break;
2532		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
 
2533			break;
2534		case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
 
2535			break;
2536		}
2537	}
2538
2539	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2540	if (start < 0)
2541		return start;
2542
2543	tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2544				     GFP_KERNEL);
2545
2546	if (!tun_dst)
2547		return -ENOMEM;
2548
2549	err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2550	if (err) {
2551		dst_release((struct dst_entry *)tun_dst);
2552		return err;
2553	}
2554
2555	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2556			 sizeof(*ovs_tun), log);
2557	if (IS_ERR(a)) {
2558		dst_release((struct dst_entry *)tun_dst);
2559		return PTR_ERR(a);
2560	}
2561
2562	ovs_tun = nla_data(a);
2563	ovs_tun->tun_dst = tun_dst;
2564
2565	tun_info = &tun_dst->u.tun_info;
2566	tun_info->mode = IP_TUNNEL_INFO_TX;
2567	if (key.tun_proto == AF_INET6)
2568		tun_info->mode |= IP_TUNNEL_INFO_IPV6;
 
 
2569	tun_info->key = key.tun_key;
2570
2571	/* We need to store the options in the action itself since
2572	 * everything else will go away after flow setup. We can append
2573	 * it to tun_info and then point there.
2574	 */
2575	ip_tunnel_info_opts_set(tun_info,
2576				TUN_METADATA_OPTS(&key, key.tun_opts_len),
2577				key.tun_opts_len);
2578	add_nested_action_end(*sfa, start);
2579
2580	return err;
2581}
2582
2583static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2584			 bool is_push_nsh, bool log)
2585{
2586	struct sw_flow_match match;
2587	struct sw_flow_key key;
2588	int ret = 0;
2589
2590	ovs_match_init(&match, &key, true, NULL);
2591	ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2592				      is_push_nsh, log);
2593	return !ret;
2594}
2595
2596/* Return false if there are any non-masked bits set.
2597 * Mask follows data immediately, before any netlink padding.
2598 */
2599static bool validate_masked(u8 *data, int len)
2600{
2601	u8 *mask = data + len;
2602
2603	while (len--)
2604		if (*data++ & ~*mask++)
2605			return false;
2606
2607	return true;
2608}
2609
2610static int validate_set(const struct nlattr *a,
2611			const struct sw_flow_key *flow_key,
2612			struct sw_flow_actions **sfa, bool *skip_copy,
2613			u8 mac_proto, __be16 eth_type, bool masked, bool log)
2614{
2615	const struct nlattr *ovs_key = nla_data(a);
2616	int key_type = nla_type(ovs_key);
2617	size_t key_len;
2618
2619	/* There can be only one key in a action */
2620	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2621		return -EINVAL;
2622
2623	key_len = nla_len(ovs_key);
2624	if (masked)
2625		key_len /= 2;
2626
2627	if (key_type > OVS_KEY_ATTR_MAX ||
2628	    !check_attr_len(key_len, ovs_key_lens[key_type].len))
2629		return -EINVAL;
2630
2631	if (masked && !validate_masked(nla_data(ovs_key), key_len))
2632		return -EINVAL;
2633
2634	switch (key_type) {
2635	const struct ovs_key_ipv4 *ipv4_key;
2636	const struct ovs_key_ipv6 *ipv6_key;
2637	int err;
2638
2639	case OVS_KEY_ATTR_PRIORITY:
2640	case OVS_KEY_ATTR_SKB_MARK:
2641	case OVS_KEY_ATTR_CT_MARK:
2642	case OVS_KEY_ATTR_CT_LABELS:
2643		break;
2644
2645	case OVS_KEY_ATTR_ETHERNET:
2646		if (mac_proto != MAC_PROTO_ETHERNET)
2647			return -EINVAL;
2648		break;
2649
2650	case OVS_KEY_ATTR_TUNNEL:
 
 
2651		if (masked)
2652			return -EINVAL; /* Masked tunnel set not supported. */
2653
2654		*skip_copy = true;
2655		err = validate_and_copy_set_tun(a, sfa, log);
2656		if (err)
2657			return err;
2658		break;
 
 
 
2659
2660	case OVS_KEY_ATTR_IPV4:
2661		if (eth_type != htons(ETH_P_IP))
2662			return -EINVAL;
2663
2664		ipv4_key = nla_data(ovs_key);
2665
2666		if (masked) {
2667			const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2668
2669			/* Non-writeable fields. */
2670			if (mask->ipv4_proto || mask->ipv4_frag)
2671				return -EINVAL;
2672		} else {
2673			if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2674				return -EINVAL;
2675
2676			if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2677				return -EINVAL;
2678		}
2679		break;
 
 
 
2680
2681	case OVS_KEY_ATTR_IPV6:
2682		if (eth_type != htons(ETH_P_IPV6))
2683			return -EINVAL;
2684
2685		ipv6_key = nla_data(ovs_key);
2686
2687		if (masked) {
2688			const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2689
2690			/* Non-writeable fields. */
2691			if (mask->ipv6_proto || mask->ipv6_frag)
2692				return -EINVAL;
2693
2694			/* Invalid bits in the flow label mask? */
2695			if (ntohl(mask->ipv6_label) & 0xFFF00000)
2696				return -EINVAL;
2697		} else {
2698			if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2699				return -EINVAL;
2700
2701			if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2702				return -EINVAL;
2703		}
2704		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2705			return -EINVAL;
2706
2707		break;
2708
2709	case OVS_KEY_ATTR_TCP:
2710		if ((eth_type != htons(ETH_P_IP) &&
2711		     eth_type != htons(ETH_P_IPV6)) ||
2712		    flow_key->ip.proto != IPPROTO_TCP)
2713			return -EINVAL;
2714
2715		break;
2716
2717	case OVS_KEY_ATTR_UDP:
2718		if ((eth_type != htons(ETH_P_IP) &&
2719		     eth_type != htons(ETH_P_IPV6)) ||
2720		    flow_key->ip.proto != IPPROTO_UDP)
2721			return -EINVAL;
2722
2723		break;
2724
2725	case OVS_KEY_ATTR_MPLS:
2726		if (!eth_p_mpls(eth_type))
2727			return -EINVAL;
2728		break;
2729
2730	case OVS_KEY_ATTR_SCTP:
2731		if ((eth_type != htons(ETH_P_IP) &&
2732		     eth_type != htons(ETH_P_IPV6)) ||
2733		    flow_key->ip.proto != IPPROTO_SCTP)
2734			return -EINVAL;
2735
2736		break;
2737
2738	case OVS_KEY_ATTR_NSH:
2739		if (eth_type != htons(ETH_P_NSH))
2740			return -EINVAL;
2741		if (!validate_nsh(nla_data(a), masked, false, log))
2742			return -EINVAL;
2743		break;
2744
2745	default:
2746		return -EINVAL;
2747	}
2748
2749	/* Convert non-masked non-tunnel set actions to masked set actions. */
2750	if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2751		int start, len = key_len * 2;
2752		struct nlattr *at;
2753
2754		*skip_copy = true;
2755
2756		start = add_nested_action_start(sfa,
2757						OVS_ACTION_ATTR_SET_TO_MASKED,
2758						log);
2759		if (start < 0)
2760			return start;
2761
2762		at = __add_action(sfa, key_type, NULL, len, log);
2763		if (IS_ERR(at))
2764			return PTR_ERR(at);
2765
2766		memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2767		memset(nla_data(at) + key_len, 0xff, key_len);    /* Mask. */
2768		/* Clear non-writeable bits from otherwise writeable fields. */
2769		if (key_type == OVS_KEY_ATTR_IPV6) {
2770			struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2771
2772			mask->ipv6_label &= htonl(0x000FFFFF);
2773		}
2774		add_nested_action_end(*sfa, start);
2775	}
2776
2777	return 0;
2778}
2779
2780static int validate_userspace(const struct nlattr *attr)
2781{
2782	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2783		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2784		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2785		[OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2786	};
2787	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2788	int error;
2789
2790	error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX, attr,
2791				 userspace_policy, NULL);
2792	if (error)
2793		return error;
2794
2795	if (!a[OVS_USERSPACE_ATTR_PID] ||
2796	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2797		return -EINVAL;
2798
2799	return 0;
2800}
2801
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2802static int copy_action(const struct nlattr *from,
2803		       struct sw_flow_actions **sfa, bool log)
2804{
2805	int totlen = NLA_ALIGN(from->nla_len);
2806	struct nlattr *to;
2807
2808	to = reserve_sfa_size(sfa, from->nla_len, log);
2809	if (IS_ERR(to))
2810		return PTR_ERR(to);
2811
2812	memcpy(to, from, totlen);
2813	return 0;
2814}
2815
2816static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2817				  const struct sw_flow_key *key,
2818				  struct sw_flow_actions **sfa,
2819				  __be16 eth_type, __be16 vlan_tci, bool log)
 
2820{
2821	u8 mac_proto = ovs_key_mac_proto(key);
2822	const struct nlattr *a;
2823	int rem, err;
2824
2825	nla_for_each_nested(a, attr, rem) {
2826		/* Expected argument lengths, (u32)-1 for variable length. */
2827		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2828			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2829			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2830			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2831			[OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2832			[OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2833			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2834			[OVS_ACTION_ATTR_POP_VLAN] = 0,
2835			[OVS_ACTION_ATTR_SET] = (u32)-1,
2836			[OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2837			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2838			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2839			[OVS_ACTION_ATTR_CT] = (u32)-1,
2840			[OVS_ACTION_ATTR_CT_CLEAR] = 0,
2841			[OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
2842			[OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
2843			[OVS_ACTION_ATTR_POP_ETH] = 0,
2844			[OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
2845			[OVS_ACTION_ATTR_POP_NSH] = 0,
2846			[OVS_ACTION_ATTR_METER] = sizeof(u32),
 
 
 
 
2847		};
2848		const struct ovs_action_push_vlan *vlan;
2849		int type = nla_type(a);
2850		bool skip_copy;
2851
2852		if (type > OVS_ACTION_ATTR_MAX ||
2853		    (action_lens[type] != nla_len(a) &&
2854		     action_lens[type] != (u32)-1))
2855			return -EINVAL;
2856
2857		skip_copy = false;
2858		switch (type) {
2859		case OVS_ACTION_ATTR_UNSPEC:
2860			return -EINVAL;
2861
2862		case OVS_ACTION_ATTR_USERSPACE:
2863			err = validate_userspace(a);
2864			if (err)
2865				return err;
2866			break;
2867
2868		case OVS_ACTION_ATTR_OUTPUT:
2869			if (nla_get_u32(a) >= DP_MAX_PORTS)
2870				return -EINVAL;
2871			break;
2872
2873		case OVS_ACTION_ATTR_TRUNC: {
2874			const struct ovs_action_trunc *trunc = nla_data(a);
2875
2876			if (trunc->max_len < ETH_HLEN)
2877				return -EINVAL;
2878			break;
2879		}
2880
2881		case OVS_ACTION_ATTR_HASH: {
2882			const struct ovs_action_hash *act_hash = nla_data(a);
2883
2884			switch (act_hash->hash_alg) {
2885			case OVS_HASH_ALG_L4:
2886				break;
2887			default:
2888				return  -EINVAL;
2889			}
2890
2891			break;
2892		}
2893
2894		case OVS_ACTION_ATTR_POP_VLAN:
2895			if (mac_proto != MAC_PROTO_ETHERNET)
2896				return -EINVAL;
2897			vlan_tci = htons(0);
2898			break;
2899
2900		case OVS_ACTION_ATTR_PUSH_VLAN:
2901			if (mac_proto != MAC_PROTO_ETHERNET)
2902				return -EINVAL;
2903			vlan = nla_data(a);
2904			if (!eth_type_vlan(vlan->vlan_tpid))
2905				return -EINVAL;
2906			if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2907				return -EINVAL;
2908			vlan_tci = vlan->vlan_tci;
2909			break;
2910
2911		case OVS_ACTION_ATTR_RECIRC:
2912			break;
2913
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2914		case OVS_ACTION_ATTR_PUSH_MPLS: {
2915			const struct ovs_action_push_mpls *mpls = nla_data(a);
2916
2917			if (!eth_p_mpls(mpls->mpls_ethertype))
2918				return -EINVAL;
2919			/* Prohibit push MPLS other than to a white list
2920			 * for packets that have a known tag order.
2921			 */
2922			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2923			    (eth_type != htons(ETH_P_IP) &&
2924			     eth_type != htons(ETH_P_IPV6) &&
2925			     eth_type != htons(ETH_P_ARP) &&
2926			     eth_type != htons(ETH_P_RARP) &&
2927			     !eth_p_mpls(eth_type)))
2928				return -EINVAL;
2929			eth_type = mpls->mpls_ethertype;
 
2930			break;
2931		}
2932
2933		case OVS_ACTION_ATTR_POP_MPLS:
2934			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
 
2935			    !eth_p_mpls(eth_type))
2936				return -EINVAL;
2937
2938			/* Disallow subsequent L2.5+ set and mpls_pop actions
2939			 * as there is no check here to ensure that the new
2940			 * eth_type is valid and thus set actions could
2941			 * write off the end of the packet or otherwise
2942			 * corrupt it.
 
2943			 *
2944			 * Support for these actions is planned using packet
2945			 * recirculation.
2946			 */
2947			eth_type = htons(0);
 
 
 
 
 
 
 
 
 
 
 
 
2948			break;
 
2949
2950		case OVS_ACTION_ATTR_SET:
2951			err = validate_set(a, key, sfa,
2952					   &skip_copy, mac_proto, eth_type,
2953					   false, log);
2954			if (err)
2955				return err;
2956			break;
2957
2958		case OVS_ACTION_ATTR_SET_MASKED:
2959			err = validate_set(a, key, sfa,
2960					   &skip_copy, mac_proto, eth_type,
2961					   true, log);
2962			if (err)
2963				return err;
2964			break;
2965
2966		case OVS_ACTION_ATTR_SAMPLE: {
2967			bool last = nla_is_last(a, rem);
2968
2969			err = validate_and_copy_sample(net, a, key, sfa,
2970						       eth_type, vlan_tci,
 
2971						       log, last);
2972			if (err)
2973				return err;
2974			skip_copy = true;
2975			break;
2976		}
2977
2978		case OVS_ACTION_ATTR_CT:
2979			err = ovs_ct_copy_action(net, a, key, sfa, log);
2980			if (err)
2981				return err;
2982			skip_copy = true;
2983			break;
2984
2985		case OVS_ACTION_ATTR_CT_CLEAR:
2986			break;
2987
2988		case OVS_ACTION_ATTR_PUSH_ETH:
2989			/* Disallow pushing an Ethernet header if one
2990			 * is already present */
2991			if (mac_proto != MAC_PROTO_NONE)
2992				return -EINVAL;
2993			mac_proto = MAC_PROTO_NONE;
2994			break;
2995
2996		case OVS_ACTION_ATTR_POP_ETH:
2997			if (mac_proto != MAC_PROTO_ETHERNET)
2998				return -EINVAL;
2999			if (vlan_tci & htons(VLAN_TAG_PRESENT))
3000				return -EINVAL;
3001			mac_proto = MAC_PROTO_ETHERNET;
3002			break;
3003
3004		case OVS_ACTION_ATTR_PUSH_NSH:
3005			if (mac_proto != MAC_PROTO_ETHERNET) {
3006				u8 next_proto;
3007
3008				next_proto = tun_p_from_eth_p(eth_type);
3009				if (!next_proto)
3010					return -EINVAL;
3011			}
3012			mac_proto = MAC_PROTO_NONE;
3013			if (!validate_nsh(nla_data(a), false, true, true))
3014				return -EINVAL;
3015			break;
3016
3017		case OVS_ACTION_ATTR_POP_NSH: {
3018			__be16 inner_proto;
3019
3020			if (eth_type != htons(ETH_P_NSH))
3021				return -EINVAL;
3022			inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3023			if (!inner_proto)
3024				return -EINVAL;
3025			if (key->nsh.base.np == TUN_P_ETHERNET)
3026				mac_proto = MAC_PROTO_ETHERNET;
3027			else
3028				mac_proto = MAC_PROTO_NONE;
3029			break;
3030		}
3031
3032		case OVS_ACTION_ATTR_METER:
3033			/* Non-existent meters are simply ignored.  */
3034			break;
3035
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3036		default:
3037			OVS_NLERR(log, "Unknown Action type %d", type);
3038			return -EINVAL;
3039		}
3040		if (!skip_copy) {
3041			err = copy_action(a, sfa, log);
3042			if (err)
3043				return err;
3044		}
3045	}
3046
3047	if (rem > 0)
3048		return -EINVAL;
3049
3050	return 0;
3051}
3052
3053/* 'key' must be the masked key. */
3054int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3055			 const struct sw_flow_key *key,
3056			 struct sw_flow_actions **sfa, bool log)
3057{
3058	int err;
 
3059
3060	*sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3061	if (IS_ERR(*sfa))
3062		return PTR_ERR(*sfa);
3063
 
 
 
3064	(*sfa)->orig_len = nla_len(attr);
3065	err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3066				     key->eth.vlan.tci, log);
3067	if (err)
3068		ovs_nla_free_flow_actions(*sfa);
3069
3070	return err;
3071}
3072
3073static int sample_action_to_attr(const struct nlattr *attr,
3074				 struct sk_buff *skb)
3075{
3076	struct nlattr *start, *ac_start = NULL, *sample_arg;
3077	int err = 0, rem = nla_len(attr);
3078	const struct sample_arg *arg;
3079	struct nlattr *actions;
3080
3081	start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
3082	if (!start)
3083		return -EMSGSIZE;
3084
3085	sample_arg = nla_data(attr);
3086	arg = nla_data(sample_arg);
3087	actions = nla_next(sample_arg, &rem);
3088
3089	if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3090		err = -EMSGSIZE;
3091		goto out;
3092	}
3093
3094	ac_start = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
3095	if (!ac_start) {
3096		err = -EMSGSIZE;
3097		goto out;
3098	}
3099
3100	err = ovs_nla_put_actions(actions, rem, skb);
3101
3102out:
3103	if (err) {
3104		nla_nest_cancel(skb, ac_start);
3105		nla_nest_cancel(skb, start);
3106	} else {
3107		nla_nest_end(skb, ac_start);
3108		nla_nest_end(skb, start);
3109	}
3110
3111	return err;
3112}
3113
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3114static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3115{
3116	const struct nlattr *ovs_key = nla_data(a);
3117	int key_type = nla_type(ovs_key);
3118	struct nlattr *start;
3119	int err;
3120
3121	switch (key_type) {
3122	case OVS_KEY_ATTR_TUNNEL_INFO: {
3123		struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3124		struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3125
3126		start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
3127		if (!start)
3128			return -EMSGSIZE;
3129
3130		err =  ip_tun_to_nlattr(skb, &tun_info->key,
3131					ip_tunnel_info_opts(tun_info),
3132					tun_info->options_len,
3133					ip_tunnel_info_af(tun_info));
3134		if (err)
3135			return err;
3136		nla_nest_end(skb, start);
3137		break;
3138	}
3139	default:
3140		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3141			return -EMSGSIZE;
3142		break;
3143	}
3144
3145	return 0;
3146}
3147
3148static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3149						struct sk_buff *skb)
3150{
3151	const struct nlattr *ovs_key = nla_data(a);
3152	struct nlattr *nla;
3153	size_t key_len = nla_len(ovs_key) / 2;
3154
3155	/* Revert the conversion we did from a non-masked set action to
3156	 * masked set action.
3157	 */
3158	nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
3159	if (!nla)
3160		return -EMSGSIZE;
3161
3162	if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3163		return -EMSGSIZE;
3164
3165	nla_nest_end(skb, nla);
3166	return 0;
3167}
3168
3169int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3170{
3171	const struct nlattr *a;
3172	int rem, err;
3173
3174	nla_for_each_attr(a, attr, len, rem) {
3175		int type = nla_type(a);
3176
3177		switch (type) {
3178		case OVS_ACTION_ATTR_SET:
3179			err = set_action_to_attr(a, skb);
3180			if (err)
3181				return err;
3182			break;
3183
3184		case OVS_ACTION_ATTR_SET_TO_MASKED:
3185			err = masked_set_action_to_set_action_attr(a, skb);
3186			if (err)
3187				return err;
3188			break;
3189
3190		case OVS_ACTION_ATTR_SAMPLE:
3191			err = sample_action_to_attr(a, skb);
3192			if (err)
3193				return err;
3194			break;
3195
3196		case OVS_ACTION_ATTR_CT:
3197			err = ovs_ct_action_to_attr(nla_data(a), skb);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3198			if (err)
3199				return err;
3200			break;
3201
3202		default:
3203			if (nla_put(skb, type, nla_len(a), nla_data(a)))
3204				return -EMSGSIZE;
3205			break;
3206		}
3207	}
3208
3209	return 0;
3210}